[Senate Hearing 117-681]
[From the U.S. Government Publishing Office]
S. Hrg. 117-681
SOCIAL MEDIA'S IMPACT ON HOMELAND SECURITY
=======================================================================
HEARING
before the
COMMITTEE ON
HOMELAND SECURITY AND GOVERNMENTAL AFFAIRS
UNITED STATES SENATE
ONE HUNDRED SEVENTEENTH CONGRESS
SECOND SESSION
----------
SEPTEMBER 14, 2022
----------
Available via the World Wide Web: http://govinfo.gov
Printed for the use of the
Committee on Homeland Security and Governmental Affairs
[GRAPHIC(S) NOT AVAILABLE IN TIFF FORMAT]
SOCIAL MEDIA'S IMPACT ON HOMELAND SECURITY
S. Hrg. 117-681
SOCIAL MEDIA'S IMPACT ON HOMELAND SECURITY
=======================================================================
HEARING
before the
COMMITTEE ON
HOMELAND SECURITY AND GOVERNMENTAL AFFAIRS
UNITED STATES SENATE
ONE HUNDRED SEVENTEENTH CONGRESS
SECOND SESSION
__________
SEPTEMBER 14, 2022
__________
Available via the World Wide Web: http://www.govinfo.gov
Printed for the use of the
Committee on Homeland Security and Governmental Affairs
GRAPHIC(S) NOT AVAILABLE IN TIFF FORMAT]
COMMITTEE ON HOMELAND SECURITY AND GOVERNMENTAL AFFAIRS
______
U.S. GOVERNMENT PUBLISHING OFFICE
51-542 WASHINGTON : 2023
GARY C. PETERS, Michigan, Chairman
THOMAS R. CARPER, Delaware ROB PORTMAN, Ohio
MAGGIE HASSAN, New Hampshire RON JOHNSON, Wisconsin
KYRSTEN SINEMA, Arizona RAND PAUL, Kentucky
JACKY ROSEN, Nevada JAMES LANKFORD, Oklahoma
ALEX PADILLA, California MITT ROMNEY, Utah
JON OSSOFF, Georgia RICK SCOTT, Florida
JOSH HAWLEY, Missouri
David M. Weinberg, Staff Director
Zachary I. Schram, Chief Counsel
Christopher J. Mulkins, Director of Homeland Security
Alan Kahn, Chief Investigative Counsel
Moran Banai, Senior Professional Staff Member
Pamela Thiessen, Minority Staff Director
Sam J. Mulopulos, Minority Deputy Staff Director
Clyde E. Hicks, Jr., Minority Director of Homeland Security
Margaret E. Frankel, Minority Professional Staff Member
Laura W. Kilbride, Chief Clerk
Ashley A. Gonzalez, Hearing Clerk
C O N T E N T S
------
Opening statements:
Page
Senator Peters
Senator Portman
Senator Johnson
Senator Lankford
Senator Romney
Senator Hawley
Senator Rosen................................................ 26
Senator Hassan............................................... 29
Senator Carper............................................... 53
Senator Sinema............................................... 58
Senator Padilla.............................................. 61
Senator Ossoff............................................... 68
Senator Scott................................................ 86
Prepared statements:
Senator Peters
Senator Portman
WITNESSES
Wednesday, September 14, 2022
Alex Roetter, Former Senior Vice President for Engineering (2014-
2016) Twitter.................................................. 5
Brian Boland, Former Vice President (2018-2020) Partnerships
Product Marketing, Partner Engineering, Marketing, Strategic
Operations & Analytics Facebook................................ 7
Geoffrey Cain, Senior Fellow for Critical Emerging Technologies,
Lincoln Network................................................ 9
Chris Cox, Chief Product Officer, META........................... 41
Neal Mohan, Chief Product Officer, YouTube....................... 42
Vanessa Pappas, Chief Operating Officer, TikTok.................. 44
Jay Sullivan, General Manager of Bluebird, Twitter............... 46
Alphabetical List of Witnesses
Boland, Brian:
Testimony.................................................... 7
Prepared statement........................................... 110
Cain, Geoffrey:
Testimony.................................................... 9
Prepared statement........................................... 117
Cox, Chris:
Testimony.................................................... 41
Prepared statement........................................... 125
Mohan, Neal:
Testimony.................................................... 42
Prepared statement........................................... 129
Pappas, Vanessa:
Testimony.................................................... 44
Prepared statement........................................... 136
Roetter, Alex:
Testimony.................................................... 5
Prepared statement........................................... 104
Sullivan, Jay:
Testimony.................................................... 46
Prepared statement........................................... 154
APPENDIX
Senator Peters Washington Post article........................... 161
Senator Peters Facebook Auto Generates Pages for Extremist Groups 162
Verge Article.................................................... 163
Senator Johnson quote from Rochelle Walensky..................... 192
Senator Johnson censored chart................................... 193
Senator Johnson Drug Adverse Event Comparison Chart.............. 194
Senator Johnson chart from Public Health England................. 195
Senator Scott chart.............................................. 196
Google's response letter to Senator Scott........................ 197
Response to post-hearing questions submitted for the Record
Mr. Cain..................................................... 200
Mr. Cox...................................................... 203
Mr. Mohan.................................................... 248
Ms. Pappas................................................... 290
Mr. Sullivan................................................. 339
SOCIAL MEDIA'S IMPACT ON HOMELAND SECURITY
----------
WEDNESDAY, SEPTEMBER 14, 2022
U.S. Senate,
Committee on Homeland Security
and Governmental Affairs,
Washington, DC.
The Committee met, pursuant to notice, at 10 a.m., in room
SD-342, Dirksen Senate Office Building, Hon. Gary Peters,
Chairman of the Committee, presiding.
Present: Senators Peters, Hassan, Rosen, Ossoff, Portman,
Johnson, Lankford, Romney, Scott, and Hawley.
OPENING STATEMENT OF CHAIRMAN PETERS\1\
Chairman Peters. The Committee will come to order.
---------------------------------------------------------------------------
\1\ The prepared statement of Senator Peters appears in the
Appendix on page 95.
---------------------------------------------------------------------------
In recent years, domestic terrorism, and specifically white
supremacist, conspiracy related, and anti-government violence,
has become one of our nation's greatest homeland security
threats.
Last October, the Committee held a hearing to examine the
role social media platforms play in the amplification of
domestic extremist content and how that content can translate
into real-world violence. We heard from expert witnesses who
discussed how recommendation algorithms, ad targeting, and
other amplification tools end up pushing increasingly extreme
content to users because that type of content is what keeps
people active on the platforms.
Unfortunately, because these platforms are designed to push
the most engaging posts to more users, they end up amplifying
extremist, dangerous and radicalizing content. This includes
QAnon, Stop the Steal, and other conspiracy theories, as well
as white supremacist and Anti-Semitic rhetoric.
In some cases, this content may not necessarily violate a
company's community guidelines. In other cases, even content
that is in clear violation of company policies remains on the
platforms, and is often only removed after public pressure. In
both cases, this content does significant harm to our society
and stokes real-world violence.
We have seen this happen time and time again. From the 2017
neo-Nazi ``Unite the Right'' rally in Charlottesville, Virginia
that was organized using a Facebook event page, to the violent
January 6, 2021, attack on the U.S. Capitol spurred to action
in part by Stop the Steal content that repeatedly surfaced
online, to the shooter who livestreamed as he massacred Black
shoppers at a Buffalo supermarket, there is a clear connection
between online content and offline violence.
Over the years, we have heard many explanations from social
media companies about their content moderation policies,
efforts to boost trust and safety, and actions taken to remove
harmful accounts.
There is no question that those efforts are certainly
important, but there is a question of whether those actions are
enough to effectively address the spread of dangerous content
online and the resulting threats it poses to our homeland
security.
The central question is not just what content the platforms
can take down once it is posted, but how they design their
products in a way that boosts this content in the first place,
and whether they build those products with safety in mind to
effectively address how harmful content spreads.
That is the focus of today's hearing where we will have the
opportunity to hear from two panels of witnesses, outside
experts, including former Facebook and Twitter executives, as
well as current senior executives from Meta, YouTube, TikTok,
and Twitter, who are charged with designing social media
products used by billions of people around the world.
The overwhelming majority of social media users have very
little information about why they see certain recommended
content in their feed, and there is very limited transparency
into how social media companies balance their business
decisions with the need for online safety, including what
resources they invest into limiting the spread of harmful
content.
Our goal is to better understand how company business
models and incentive structures, including revenue generation,
growth, and employee compensation, determine how social media
products are built and the extent to which current incentives
contribute to the amplification of content that threatens
homeland security.
For nearly a year, I have been pressing Meta, YouTube,
TikTok, and Twitter for more information on their policies to
monitor and remove extremist and conspiracy content that
advocates violence, as well as the relationship between their
recommendation algorithms and targeted advertising tools that
generate much of the companies' revenues, and the amplification
of extremist content.
The companies' response to those inquiries have been
incomplete and insufficient so far.
This morning, we will hear from two former executives and a
technology journalist with social media expertise about the
internal product development process and the business decisions
these companies make, including tradeoffs between revenues and
growth and their trust and safety efforts, as well as how they
interact with foreign governments.
Later this afternoon we will hear directly from the Chief
Product Officers (CPO) of Meta, YouTube, and Twitter and the
Chief Operating Officer (COO) of TikTok, the Executives who are
charged with making these business decisions and driving the
strategic vision of the companies.
I certainly look forward to a productive discussion with
both panelists. Welcome to this Committee here today. We look
forward to your testimony.
Ranking Member Portman, you are now recognized for your
opening comments.
OPENING STATEMENT OF SENATOR PORTMAN\1\
Senator Portman. Thank you, Mr. Chairman, and I thank the
experts for being here. We look forward to hearing from you.
This is going to be an interesting hearing.
---------------------------------------------------------------------------
\1\ The prepared statement of Senator Portman appears in the
Appendix on page 99.
---------------------------------------------------------------------------
This past Sunday observed the 21st anniversary of the
tragic September 11, 2001 (9/11) terrorist attacks, and over
these past couple of decades our country has adapted to combat
the most pressing threats to our nation's security, and that is
good. But the advent of social media has added a new dimension
to the ever-evolving threat landscape and created new
considerations for combating terrorism, human trafficking, and
many other threats.
During last October's hearing on how algorithms promote
harmful content I focused on how social media business models
contribute to the amplification of terrorism and other
dangerous activities. Since then, the Committee has identified
ways in which social media companies' product development
processes tend to conflict with user safety. Whistleblower
testimony has revealed that in numerous occasions the leaders
at social media companies were aware that certain platform
features increased threats to user safety and chose not to
mitigate such concerns. We will hear about that today.
It is unfortunate that the American public must wait for
whistleblower disclosures to find out about ways in which
platforms are knowingly and unknowingly harming their users.
The lack of transparency in the product development process,
the obscurity of algorithms, and misleading content moderation
statistics create an asymmetric information environment in
which the platforms know all, yet the users and policymakers
and the public actually know very little.
One consequence of this lack of transparency is related to
China. I have serious concerns about the opportunities that the
Chinese Communist Party (CCP) has to access TikTok's data on
American users. There are now over 100 million Americans,
including 40 million under the age of 19 who use TikTok. This
TikTok data remains vulnerable to the Communist Party of China,
both as the CCP tries to exploit its access to U.S. data and
exert influence over the content that U.S. users see.
For example, despite moving U.S. data servers to the United
States, TikTok and ByteDance employees in China retain the
ability to access this data. If that is not true we would like
to hear about that today.
Also we learned yesterday, from Senator Grassley's opening
statement in a Senate Judiciary Committee hearing with the
Twitter whistleblower that Twitter failed to prevent Americans'
data from being accessed by foreign governments. In fact,
Senator Grassley spoke about how several Twitter employees were
actually foreign agents of India, China, and Saudi Arabia,
which is concerning and speaks to why Congress needs more
information from platforms on how they secure user data.
Another consequence of poor transparency relates to content
moderation. While I recognize that content moderation is a key
component to creating safe platforms for users, it cannot be
the only thing. Transparency reports released by companies
often detail the amount of content that has been removed for
violating company policy. However, these reports do not account
for violating content that is left up on the platform and yet
goes undetected.
It also does not account for content that is incorrectly
censored, as we often see with many conservative voices on
social media. I, like many of my colleagues, have been critical
of the political biases held by big tech platforms, which have
resulted in systematic takedowns of accounts that hold
ideologies with which the left and liberal media disagree.
We will hear about that today, but these takedowns are
often done under the guise of combating misinformation which,
in fact, they are just combating conservative viewpoints that
conflict with their own. Any steps taken to address the impact
media on homeland security must account for First Amendment
protections, of course, and safeguard free speech.
For us to have a responsible conversation about the impact
of harmful content on American users and homeland security we
need to talk about how current transparency efforts have worked
or not worked. Congress must enact legislation that will
require tech companies to share necessary data so that research
can be done to evaluate the true extent of how harms from
social media impact Americans.
As some of you know, I have been working on legislation
along those lines with Senator Coons to establish bipartisan
legislation to do just that. The Platform Accountability and
Transparency Act (PATA) would require the largest tech
platforms to share data with vetted independent researchers and
other investigators so that we can all increase our
understanding of the inner workings of social media companies
and regulate the industry based on good information that we
simply do not have now, that we can learn through this process.
Again, I thank the witnesses for being here and I look
forward to having your expertise help to guide us in these
complicated issues, and thank you, Mr. Chairman, for holding
this hearing.
Chairman Peters. Thank you, Ranking Member Portman.
It is the practice of this Homeland Security and
Governmental Affairs Committee (HSGAC) to swear in witnesses,
so if each of you would please stand and raise your right
hands.
Do you swear that the testimony you will give before this
Committee will be the truth, the whole truth, and nothing but
the truth, so help you, God?
Mr. Roetter. I do.
Mr. Boland. I do.
Mr. Cain. I do.
Chairman Peters. You may be seated.
Today's first witness is Alex Roetter, the former Senior
Vice President of Engineering at Twitter. In his previous role,
Mr. Roetter helped grow Twitter's monthly active users to over
300 million and build the ad network from near zero revenue to
$2.5 billion a year.
Mr. Roetter also spent six years at Google on a variety of
projects including building the world's largest computational
advertising platform. He was in the room for major decisions
about products at Twitter and is familiar with the priorities
that were weighed as products were created, as well as how
those products are then built.
Mr. Roetter, welcome to the Committee. You may proceed with
your opening remarks.
TESTIMONY OF ALEX ROETTER,\1\ FORMER SENIOR VICE PRESIDENT FOR
ENGINEERING (2014-2016), TWITTER
Mr. Roetter. Good morning, Mr. Chairman and members of the
Committee. Thank you for inviting me here today.
---------------------------------------------------------------------------
\1\ The prepared statement of Mr. Roetter appears in the Appendix
on page 104.
---------------------------------------------------------------------------
We live in a world where an unprecedented number of people
consume information from social networks. Viral content and
misinformation can propagate on these platforms on a scale that
is unseen in human history. Regulators must understand
companies' incentives, culture, and processes to appreciate how
unlikely voluntary reform is.
In over 20 years of working in Silicon Valley as an
engineer and as an Executive, I have seen firsthand how several
of these companies work. Today I will talk about how these
companies operate and actionable ways to demand transparency.
The product development lifecycle works as follows. First,
teams of product managers, engineers, and designers are
assigned specific metrics to maximize. These metrics carefully
track user engagement and growth as well as revenue and
financial indicators. Other metrics, such as user safety, are
either not present or much less important.
Second, teams use an experimental system to launch changes
to small percentages of users. The effect of every experiment
on key metrics is measured extremely accurately. Absent are
detailed metrics tracking impacts on user safety. For example,
I never once saw a measurement such as did a given experiment
increase or decrease the spread of content later identified as
hate speech.
Third, executives review these experimental dashboards
regularly and make decisions on which experiments to launch.
These reviews are run by product and engineering. Other
functions like legal or trust and safety are absent or do not
play a substantial role.
Culturally, these companies are informal hierarchies with
the ``builders,'' by which I mean engineers, product managers,
and designers, held in the highest regard. Other functions are
viewed much more skeptically. The strong bias is to make sure
that corporate bureaucracy does not slow down product
development.
These companies conduct regular performance evaluations and
promotions, and these drive peer recognition, career
advancement, and cash and stock awards. The main data collect
is what impact an individual's work has on key metric families.
Only a minority of builders get promoted based on impact to
trust and safety metrics, as those impacts are not valued as
highly.
What data has been shared selectively to date is mostly
non-illuminating statistics designed to create the appearance
that they are taking the problem seriously. When one of the
largest companies in the world says it is spending what seems
like a large, absolute number, that number must be put in
context and compared to the size of other initiatives, for
example, product efforts or how much they spent on stock
buybacks. Large investments amounts are not sufficient. We must
demand transparency based on measuring actual results.
Similarly, when a company points to how much content it has
taken down, that has to be understood in terms of its reach in
the network. Removing a billboard in Wyoming is very different
than removing a billboard in Times Square.
For real transparency I recommend assembling an independent
group of researchers and data scientists. Task them with
enumerating the right questions to ask and the set of data they
need to answer them. Fund them to continually do this work and
refine their questions and data requests.
The government is able to demand transparency in
technically demanding fields. For example, third-party auditors
of public company financial statements are able to balance the
public's need for reliable financial statements with a
company's need to keep information confidential.
Until such transparency exists, every assurance by any of
these companies has to be taken on faith. Transparency is
necessary but not sufficient. Until we change the fact that
user attention and profits are what companies care about above
all else, all the data-sharing in the world will not address
the problem.
Policy and legal experts have previously testified before
the Committee on ways that incentives could be changed.
Incentives matter. Companies behave differently when they care
about the quality of content. For example, having inappropriate
ads could materially harm financial performance, so most
advertising systems place ad copy removal as a step that has to
occur before the new ad ever makes its way to users. On the
other hand, user-generated content is allowed to go live
instantly.
Incentives also shape companies' recommendation algorithms.
For example, TikTok and ByteDance feed young people in China a
diet of educational science and math content via their
recommendation algorithms. The Chinese version of the app even
enforces a daily usage limit. Contrast this to how U.S.
companies target content to young Americans, optimizing their
engagement of revenue at any cost.
Any suggestion for more useful transparency will be met
with many objections. The status quo is simply too lucrative.
Do not underestimate these companies' ability to fight requests
for information. After all, the legal team at Google alone has
the same number of lawyers as all the employees of the Federal
Trade Commission (FTC).
Given what we know about companies' incentives, processes,
and culture, we should not expect meaningful progress
voluntarily, and we should view their commitments extremely
skeptically, however, with the proper transparency and
regulatory environment I believe we can change their incentives
and start to see real, measurable progress against these
problems. Thank you.
Chairman Peters. Thank you.
Our next witness is Brian Boland, a former Vice President
of Partnerships Product Marketing, Partner Engineering,
Marketing, Strategic Operations, and Analytics at Facebook. Mr.
Boland worked at Facebook for 11 years. He worked in several
roles including leading a 500-person multifunction team focused
on product strategy, market strategy, partner engineering,
operations, analytics, and marketing. These high-impact teams
worked across Facebook products and features including watch,
video, news, group admins, developers, payments, and audience
network. Before joining Facebook, he worked at Microsoft and
other tech companies.
Mr. Boland, welcome to the Committee. You may proceed with
your opening remarks.
TESTIMONY OF BRIAN BOLAND,\1\ FORMER VICE PRESIDENT (2018-
2020), PARTNERSHIPS PRODUCT MARKETING, PARTNER ENGINEERING,
MARKETING, STRATEGIC OPERATIONS, & ANALYTICS, FACEBOOK
Mr. Boland. Good morning, Mr. Chairman, and Members of the
Committee. Thank you for holding these hearings that cover such
important issues for our nation and the world, and thank you
for inviting me here today to provide testimony on my
experiences as senior executive at Facebook, now known as Meta.
---------------------------------------------------------------------------
\1\ The prepared statement of Mr. Boland appears in the Appendix on
page 110.
---------------------------------------------------------------------------
For the last few years I have grown increasingly concerned
about the roles that Facebook, Instagram, YouTube, Twitter, and
TikTok play in driving the growth of misinformation, extremism,
and generally harmful content. I worked at Facebook for 11
years in a variety of leadership roles, helping to shape
product and market strategies for a broad array of products,
including advertising news, video media, and more. During my
tenure at the company I worked for the most senior executive
and was deeply embedded in the product development process.
In my last two years of my time at the company, the
CrowdTangle team and product was a part of my organization.
CrowdTangle is a tool that provides limited, albeit industry-
leading transparency in the public news feed content on
Facebook. What finally convinced me that it was time to leave
was that despite growing evidence that the news feed may be
causing harm globally, the focus on and investments in safety
remained small and siloed.
The documents released by Frances Haugen, the Facebook
whistleblower who last fall testified here, highlight issues
around polarization globally and the power of Facebook to lead
people down a path to more extreme beliefs. These papers
demonstrate thoughtful, well-researched documentation of the
harms that concerned me. The research was done by highly
skilled Facebook employees who are experts in their field, and
was extensive.
Rather than address the serious issues raised by its own
research, Meta leadership chooses growing the company over
keeping people safe. While the company has made investments in
safety, these investments are small are routinely abandoned if
they do not impact company growth. My experience at Facebook
was that rather than seeking to research and discover issues on
the platform before others found them, they would rather
reactively work to mitigate the public relations (PR) damage
for issues that came to light.
I have come to believe that several circumstances have put
Americans at risk from the content on these platforms. The
first is the growth over safety incentive structure that leads
to products that are designed and built without a primary focus
on safety. The next is the unprecedented lack of transparency
available from these platforms so that we can analyze content
and understand the impact from these tools. Finally, the lack
of clear oversight for the business practices of these
companies.
We have faced challenges like this before with new
technologies. In the 1960s, Congress addressed the dramatic
rise in fatalities caused by the rapid increase in automobile
use in the United States. That industry experienced explosive
growth in the companies focused on growth in sales, and it
turns out safety did not sell. The creation of the National
Highway Traffic Safety Administration (NHTSA), at the time the
National Highway Safety Bureau (NHSB), empowered an agency to
study the available data, and in partnership with researchers
and other agencies to take steps to make driving in America
rapidly and significantly safer. Today, automobile
manufacturers portray safety as a selling point such that they
welcome verification of these efforts.
The problem with these social media platforms today is that
we lack public data to understand the current issues, and there
is extremely limited ability to research these platforms and
almost no ability to protect our future and creation a version
of crash testing the car. Imagine if, in the 1960s, we had no
way of knowing the deaths that were happening from cars, and we
had no way of knowing it was increasing so rapidly. That lack
of data is where we are today with social media platforms.
The reality is that for all the debate about whether social
media is predominantly good or bad, the truth is that we do not
really know. If anyone tells you they know, they do not know. I
believe that we have a right to know. The good news is that
with the right incentives in place and rules around
transparency we can develop a better understanding of these
issues and take steps to mitigate the harms.
If we take these steps we can do now what we did with the
automobile. We can empower agencies and researchers to deeply
understand the issues, and through changes in incentives,
public education, and better development, build a path to a
future where we still get the amazing benefits from these
products while mitigating the harms that we barely understand
today.
Today I hope to shed light on product development process,
internal and external incentive structures for these
organizations, and the critical importance of transparency. I
appreciate your work to better understand these issues and
deliver real-world solutions to the American people. Thank you.
Chairman Peters. Thank you.
Our final witness of our first panel is Geoffrey Cain,
Senior Fellow for Critical Emerging Technologies at Lincoln
Network. Mr. Cain is an award-winning foreign correspondent and
author. His work has taken him to the world's most
authoritarian and remote places, from inside North Korea to the
Trans-Siberian Railway across Russia, from investigations into
genocide in Cambodia to experiments in technological
surveillance in China.
Mr. Cain has served as a tech Congress fellow with the
House Foreign Affairs Committee minority and supported a range
of issues, including China, tech sanctions, and investigative
work.
Mr. Cain, welcome to our Committee. You may proceed with
your opening comments.
TESTIMONY OF GEOFFREY CAIN,\1\ SENIOR FELLOW FOR CRITICAL
EMERGING TECHNOLOGIES, LINCOLN NETWORK
Mr. Cain. Good morning, Chairman Peters, Ranking Member
Portman, and Members of the Committee. It is an honor to be
invited to testify here today on social media's impact on
national security. Today I will talk about one of the greatest
technological threats facing our homeland security and
democracy: TikTok, the social media app that reports to a
nefarious Chinese company called ByteDance.
---------------------------------------------------------------------------
\1\ The prepared statement of Mr. Cain appears in the Appendix on
page 117.
---------------------------------------------------------------------------
As an investigative journalist in China and East Asia for
13 years, I have been detained, harassed, and threatened for my
reporting on Chinese technology companies. Today I will show
you how TikTok has orchestrated a campaign of distraction and
deflection to mask the alarming truth.
Americans face the grave unprecedented threat of software
in our pockets that contains powerful surveillance and data-
gathering capabilities, owned by private companies that must
comply with the dictates of a foreign authoritarian government
ruled under the Chinese Communist Party.
The CCP had signaled its ambitions to assert global
jurisdiction over private companies everywhere as a condition
for doing business in China. TikTok, therefore, is a disaster
waiting to happen for our security and the privacy of our
citizens.
We will have TikTok executive here today later. According
to their internal public relations guidelines, leaked to the
media, they are required to, ``Downplay the parent company,
ByteDance, downplay the China association, and downplay
artificial intelligence (AI).''
The public relations guideline states that if you ask them
about the influence of the Chinese company, ByteDance, and its
influence over its American product, TikTok, which is used by
many Generation Z teenagers, its executives must deceptively
tell you that ByteDance is a separate company in China and that
you should talk to ByteDance instead.
They will attempt to confuse you, claiming that TikTok
takes a localized approach, hiring local moderators,
implementing local policies, and showing local content. They
will not tell you about an individual who is unnamed so far,
called the ``Master Admin'' in Beijing--this has been leaked to
the media, to Buzzfeed--who has had access to all Americans'
data. They also will not tell you that they, at TikTok, report
to ByteDance executives in China, and ByteDance reports to the
Chinese Communist Party.
TikTok's fast expansion into the American market was only
possible because China has rigged the market. The Chinese
government offered ByteDance vast market protection in China,
all while banning competing American social media apps,
Facebook, Instagram, Twitter, and Google.
Like all Chinese companies, ByteDance runs an in-house
Communist Party Committee that enforces the political loyalty
of its employees. In 2018, ByteDance and TikTok's founder and
previous Chief Executive Officer (CEO), a man named Zhang
Yiming, wrote a public letter promising Chinese regulators that
his company would follow ``core socialist values,'' would
introduce these ``correct values into technology and products''
and would ensure his products promoted the Chinese Communist
Party's agenda.
These values, he wrote, included ``strengthening the work
of the party construction,'' ``deepening cooperation with
official party media,'' and strengthening ``content review in
line with these party values.''
ByteDance's public statement in China should be cause for
alarm, considering American government employees, military
personnel, and workers in sensitive and strategic industries
use TikTok.
When TikTok began growing its present in the United States
in 2016 and 2017, I was an investigative journalist in China's
western region of Xinjiang, where I was writing my second book,
The Perfect Police State, which is an investigation into the
Chinese surveillance dystopia.
I learned that ByteDance and TikTok were expanding into
America, and I knew that this was ominous because I had been
speaking to a former worker for the Ministry of State Security,
a major intelligence and extremely powerful intelligence body,
who had told me that he had worked with numerous companies,
including ByteDance, to expose the data of ethnic minorities in
China. It was not hard. It simply happened.
ByteDance has also had an active role in suppressing news
about the atrocities, which included physical and psychological
torture, internment in concentration camps, forced
sterilizations, and the wholesale destructions of mosques and
other cultural artifacts. So this is very serious.
I am aware of time so I do have much more in the written
testimony if you would like to ask. But thank you for your time
today, and I look forward to answering your questions.
Chairman Peters. Thank you, Mr. Cain.
Extremist groups, including QAnon followers, Islamic State
of Iraq and Syria (ISIS), and white supremacists certainly have
expanded their ranks by recruiting individuals on major social
media platforms. The Christchurch shooter, who killed 51 people
and inspired the Poway and El Paso shooters was radicalized on
YouTube and livestreamed his attacks on Facebook to rally
others to his cause.
Three years later, a shooter in Buffalo, New York, streamed
his attack on Twitch, which acted quickly to take it down but
the video was soon circulating widely on Facebook.
Mr. Roetter, would you tell this Committee why do these
platforms' recommendation algorithms spread this extremist
content just so rapidly?
Mr. Roetter. Thank you for the question. The way to
understand these recommendations, they do not have
intentionality about specific types of content. But the way
they work is they assemble a massive amount of information,
they model everything about your usage, your interests, your
geography, who you are connected to, what you have engaged with
historically--and then they model the content and they try to
match those optimally.
What makes this so dangerous is there is a positive
feedback loop, and if you pick something not controversial,
just pick a hobby--knitting, for example--if I think that you
are somewhat into knitting I might recommend some knitting
content to you because I believe you will engage with it. You
do engage with it, and that both makes you more interested in
knitting because you are doing the hobby more, but also it
feeds back to the algorithm, which then has signaled that you
do like more of this content.
The next day, or the next session, it is more confident
that you will engage in this content and you will go further
down the rabbit hole. Obviously, with knitting that is fine,
but this is true for all sorts of content. Because of this
feedback loop, if you have some proclivity or some interest in
some topic, you will be fed more of that. That generally feeds
your interest, and you are fed more and more.
That is why we see people that start off with more things
in common than differences sort of splitting and fracturing as
they each go into their worlds that are more and more different
and have less in common with other people. This is all an
inevitable consequence of this optimization of driving
engagement.
Chairman Peters. In terms of content that takes you down
the rabbit hole, are companies able to change some of those
algorithms to prevent that from occurring, at least with that
kind of content, and how would that work?
Mr. Roetter. You certainly could, in theory. It will never
happen, given the current incentive structure. These are for-
profit companies. They are incented to maximize profit, and
before they have realizable profits they are incented to show
massive user growth to convince investors that they will be
massively profitable in the future.
The way they do that is getting people to come back to
their platform over and over. The way they do that is for
optimizing engagement. As long as the algorithms are optimizing
for showing you things that you will engage with we will always
have this positive feedback loop property. I show you something
you are interested in, you get more interested it, you are more
likely to keep engaging.
You can build an AI to train for anything, but you pick an
incentive based on the overarching incentive of the environment
you find yourself in, in this case a public company that is
reporting to shareholders. Until those incentives change we
should not expect the AI to optimize for anything other than
engagement and profit maximization.
Chairman Peters. We are going to hear later this afternoon
from chief product officers at some of these major companies.
Is it possible for them to set different priorities for product
development to address the spread of extremist content? Is that
within their purview, and is that something they should be able
to talk about?
Mr. Roetter. In practice it is not possible, and the reason
is these are just individuals. This is not a matter of a few
bad eggs running companies. This is a system that these people
find themselves in. They are in a system where they have to
report user growth, engagement, increasing attention from the
users, and profit.
I should add, this attention game is a race to the bottom.
If I build a product that less addictive than a competitor's
product, by definition user eyeballs and attention will go over
to the competitor. Then I have to, in turn, make my product
more addictive to pull people back or I will quickly be
abandoned by investors.
Given that structure there is no way that a product leader
or any other executive at a company could optimize for anything
other than those core metrics, engagement and revenue, because
that is the system they find themselves in.
Chairman Peters. They cannot do it by themselves. It has to
be broader than that. But they are at the front end of that, or
at the beginning of that, to understand exactly how that
incentive structure, how those priorities shape the work that
they do? They can talk about that, I suspect.
Mr. Roetter. They could talk about that. They are doing
exactly what you would expect them to do, given the environment
they find themselves in. As long as the incentives of those
companies are what they are, they will continue to behave the
way they are behaving, and if they did not, it would hurt the
trajectory of the company.
Chairman Peters. Very good. Mr. Boland, why are the actions
taken by trust and security teams at these platforms just not
enough to deal with this problem? We are going to hear a lot
about these teams, I think, this afternoon. Why is that not
enough?
Mr. Boland. Yes, I imagine that you will hear that there
are significant investments from the companies in trust and
safety, and it is true they make some investments.
The important thing to think about with trust and safety
effort is that if it is siloed from the rest of the process, if
it is a last-check safeguard or a group that is small and not a
core part of the way that people build products, it will always
be an afterthought, and it will always be a team that has to
fight in the battle of tradeoffs between their ways that they
would like to improve trust and safety and impacts to growth.
Impacts to growth translate to impacts to revenue.
That dual tension of not receiving enough resources and
being at odds with the product development teams and the
product development process makes it so that team has to fight
for any sort of interventions they want to put in place.
You can change the way that product teams could work with
those organizations. A good example would be the efforts that
Facebook is now putting in place around privacy. For a number
of years we know that Facebook and Meta had not been at the top
of its game on privacy. After the last sort of issues with the
FTC, the company has invested significantly in efforts around
privacy, and has made that something that the product teams and
the product managers actually have to care about.
As long as that team in trust and safety is off to the
side, if there are not the incentives in place that say to the
company, ``You really have to make sure that when people are
making day-to-day decisions they are prioritizing these
efforts,'' that team is fighting a losing battle, both in
resources, because they are battling against a number of
people, and incentives at the company because they have to
justify every single change that they want to make. It is not a
core part of how the teams think.
Chairman Peters. At best, there is going to be follow-up.
The products are going to get launched. They are going to
potentially cause problems that team may have even recognized
but they were not able to interject that effectively during the
product development phase. Then later they may be engaged but
at that point the genie is out of the bottle, so to speak,
before they can get engaged. Is that correct?
Mr. Boland. That is correct. You can kind of understand
this if you think about where these companies started and the
incredibly short period of time that they have grown to be as
successful as they are. They still feel like startups, in the
way that the leaders think, even though these are some of the
biggest companies in the world.
At the beginning of their lifecycle it was about trying to
figure out products that could grow, and grow effectively in
the world, and has gotten to a point where they have not
matured out of that stage.
I think we can get to a point where these companies could
do more in that space. We just have not seen them make that
transition into a more responsible set of activities.
Chairman Peters. All right. Thank you.
Ranking Member Portman, you are recognized for your
questions.
Senator Portman. Thank you, Mr. Chairman. We have a good
group of Members here so I will keep within the time, and I
want to focus on TikTok first. We talked about TikTok being the
most popular social media app in the United States. I also
think it poses a risk to our national security, and I want to
dig deeper into that today with both of these panels, this one
and then when the TikTok representatives are here later.
My understanding is that under Chinese law the Chinese
Communist Party can access data of tech companies that are run
out of China or have parent companies that are run out of
China. Both ByteDance and TikTok have offices and they have
employees in Beijing.
Mr. Cain, under Chinese law does TikTok have a legal
obligation to give U.S. user data to the Chinese Communist
Party?
Mr. Cain. Oh yes, absolutely. TikTok executives will, under
Chinese law, face a minimum of 20 days' detention if they
refuse to turn over data on anyone in the world, and this could
be anybody in China, anybody who is traveling through China,
through Hong Kong. This is a documented legal situation, and it
is not something that TikTok, despite claiming to be an
American company, can avoid.
I would also like to point out that TikTok does dodge this
question frequently by trying to point out that it is run by a
Cayman Islands holding company, a shell company essentially.
This is a red herring to distract from the issue at hand.
The American company, TikTok and the Chinese company,
ByteDance, both report to this Cayman Islands shell company.
The company has never said how many people actually work for
the shell company, the holding company, but we do know that the
CEO of ByteDance and the CEO of TikTok are the same person.
This is listed on the Cayman Islands registry. The CEO is the
same person running the ByteDance company in China, according
to their website.
Senator Portman. Let me delve a little deeper here because
we are going to hear from TikTok later, and based on the
testimony we have received from them in advance I think they
are going to say they have not provided data to the Chinese
Communist Party. Even if CCP requested data, they said they
would not share it with them. Again, does China need to make a
request to access this data, or does China have the capability
to access it at will?
Mr. Cain. I am not aware of the Chinese government having
the ability to simply open a computer and access it at will. It
would usually happen through somebody in ByteDance or in
TikTok. This has already been demonstrated and documented.
There was a Buzzfeed report that came out a few months ago
which contained 20 leaked audio files from internal meeting at
TikTok in which TikTok employees said that they had to go
through Chinese employees to understand how American data was
being shared. It also pointed out that employees were saying
that there is an individual in Beijing who is called the
``Master Administrator.'' We do not know who that is yet. But
this person, according to them, had access to all data in the
TikTok system.
When they say that this data is being kept separate, it is
simply a point that has been disproven already, because we have
documentation that shows that the data has been shared
extensively.
Senator Portman. OK. We will get a chance to talk to TikTok
about that, but I appreciate your work on this and your
testimony today. We have cause for us to legislate more in this
area, generally. We talked about that earlier, regulations,
legislation.
My concern is that we really do not know what is behind the
curtain, the black box, so to speak. We proposed this
legislation called the Platform Accountability and Transparency
Act to require the largest tech platforms to share data, again
with vetted, independent researchers and other investigators.
We know what is happening with regard to user privacy, content
moderation, product development. We talked about the bias that
I believe is out there in social media today, in many of the
companies, and other practices.
Mr. Boland, you talked a little bit about this in your
testimony. I see in your written testimony you said, ``To solve
the problem of transparency we must require platforms to move
beyond the black box with legislation like the Platform
Transparency and Accountability Act.'' Can you explain why that
legislation is needed and how it would be used?
Mr. Boland. Thank you, Senator. Yes, I believe the Platform
Accountability and Transparency Act is one of the most
important pieces of legislation that is before you all. It is
not sufficient because we have to address the incentives that
we have been talking about.
But to begin with, we are at a point where we are supposed
to trust what the companies are telling us, and the companies
are telling us very little. I think Facebook, to their credit,
is telling us the most, but it kind of like a grade of a D out
of an A through F grading system. They are not telling us much,
but they are telling us more than everybody else, especially
YouTube and TikTok.
In order to understand the issues that we are concerned
about with hate speech and the way that these algorithms can
influence people, we need to have a public understanding and a
public accountability of what happens on these platforms.
There are two parts of transparency that are very
important. One is understanding what happens with moderation,
so what are the active decisions that companies are taking to
remove content or make decisions around content. There is
another critically important part that is around what are the
decisions that the algorithms built by these companies are
taking to distribute content to people.
If you have companies reporting you what they would like
to, and I am sure you will hear from them this afternoon, a lot
of averages, a lot of numbers that kind of gloss over the
concerns, if you look at averages across these large
populations you miss the story.
If you think about 220-some-odd million Americans who are
on Facebook, if one percent of them is receiving an extremely
hate-filled feed or radicalizing feed, that is over two million
people who are receiving really problematic content. In the
types of data that you are hearing today, that you are
receiving today, you get an average, which is incredibly
unhelpful.
By empowering researchers to help us understand the problem
we can do a couple of things. One, we can help the platforms,
because today they are making the decisions on their own, and I
believe that these are decisions that should be influenced by
the public. Two, then you can bring additional accountability
through an organization that has clear oversight over these
platforms. Whether that be through new rules or new fines that
you levy against the companies, you have the ability to
understand how to direct them.
Today, you do not know what is happening in the platforms.
You have to trust the companies. I lost my trust with the
companies, of what they were doing, and what Meta was doing. I
think we should move down trust to helping our researchers and
journalists understand the platforms better.
Senator Portman. OK. To the other two quickly, Mr. Cain and
Mr. Roetter, do you disagree with anything that was said about
the need for more transparency? Just quickly. I have very
little time.
Mr. Cain. I 100 percent agree.
Mr. Roetter. I 100 percent agree, and I think that this
Committee is uniquely poised, given its subpoena powers, to
enforce transparency.
Senator Portman. OK. I will have additional questions
later. Again, we have so many Members here I want to respect
the time, but I appreciate your testimony.
Chairman Peters. Thank you, Ranking Member Portman.
Senator Johnson, you are recognized for your questions.
OPENING STATEMENT OF SENATOR JOHNSON
Senator Johnson. Thank you, Mr. Chairman.
Listen, I think we all agree this is a big problem. It is
my definition of a problem, that there is no easy solution. As
Chairman of this Committee I would meet with Facebook, and I
appreciated what they were trying to do to hold down Islamic
terror type of content. I think we all agree that we do not
want to be disseminating extremist, violence-inducing type of
behavior through these platforms, but we also need to protect
free speech as well. It is a real tension. It is a real
balancing act.
Mr. Boland, I think you talked about the term extremism and
harmful content, but I guess that is all in the eye of the
beholder, is it not? It is difficult to define.
I guess what I want to focus on a little bit is, where do
we draw the line in terms of taking down content that we all
would agree is extreme and could induce violence, versus
censoring legitimate political debate?
Mr. Roetter, do you have any idea of what percentage of
Twitter employees are conservative versus liberal?
Mr. Roetter. I have no idea.
Senator Johnson. You think it is probably pretty heavily
tiled to the left. Correct?
Mr. Roetter. I do not know.
Senator Johnson. I think you do.
Mr. Boland, would you want to answer that question?
Mr. Boland. I do not know.
Senator Johnson. OK. Mr. Cain.
Mr. Cain. It is obvious, but my only knowledge is China
tech and TikTok. I am not as familiar with that area.
Senator Johnson. OK. Let me move on. Let me use an example
that I think we are all aware of, the 800-pound gorilla in the
room. Let us talk about the Hunter Biden laptop. Mr. Roetter,
do you believe that, like The Washington Post, that there is
authentic information on that laptop?
Mr. Roetter. I am not sure. I will say that these are
massive platforms. There are billions of people.
Senator Johnson. I have very little time.
Mr. Roetter. I do not know.
Senator Johnson. Mr. Boland, do you assume that is
authentic information on the laptop?
Mr. Boland. I do not have an opinion on the laptop.
Senator Johnson. OK. Twitter was actually very effective
when they blocked the New York Post articles on the Hunter
Biden laptop. We had Jack Dorsey in front of the Senate
Commerce Committee back in, I think, October 2020, and Senator
Cruz and myself asked him, because we were talking about
Russians using the platforms to impact our elections, and
everybody agrees that could happen.
We asked Mr. Dorsey, ``Do you believe Twitter could impact
the election?'' Mr. Dorsey said, ``No.'' Mr. Roetter, do you
believe Twitter has the capability of impacting an election?
Mr. Roetter. I think all of these social platforms, they
are so massive it is hard to believe that they are not
impacting.
Senator Johnson. Mr. Boland, do you believe that as well?
Mr. Boland. Yes, these platforms absolutely have influence.
Senator Johnson. Mr. Cain.
Mr. Cain. Absolutely.
Senator Johnson. OK. There is a problem right there, OK,
and I appreciate you acknowledging that fact. We had 51 former
intelligence officials. I have no idea on what basis they wrote
this letter, that came out immediately. I think it might be
because the Federal Bureau of Investigation (FBI) had a scheme
in August 2020, to downplay the derogatory information on the
Hunter Biden laptop. But they came out and said that the laptop
had all the earmarks of a Russian information operation. It
seems to me like that letter itself was an information
operation.
We have the platform censored that and Facebook throttled
it back. We actually took a poll on this. I did not but a
company called Media Research Center poll--this was after the
election--1,750 voters in seven swing States, of Biden voters
who were unaware of the emails, texts, testimony, banking
transactions on the laptop, as well as Senator Grassley's and
my report, which was based on interviews with U.S. persons and
U.S. documents.
Seventy nine percent of those Biden voters said they would
still vote for him, but 16 percent said they would not, four
percent said they would switch their vote to President Trump,
four percent would vote for a third party, four percent would
skip voting altogether, five percent would not have voted at
all. Pretty strong evidence that what Facebook and Twitter did
impacted the 2020 election to a far greater extent than
anything Russia ever could have hoped to do in 2016 or 2020.
I want to talk about other disinformation coming out of
this Committee. A day or two after Senator Grassley and I
issued our report, based on U.S. documents and interviews with
U.S. persons, our now Committee Chairman, who was then
Committee Ranking Member, issued a press release. It said,
``Peters response to a Republican effort to amplify Russian
disinformation.'' He said, ``I generated a partisan political
report that is rooted in Russian disinformation.''
Mr. Chairman, do you want to retract that false allegation
now, now that we know that the Hunter Biden laptop is accurate,
that there has not been one scintilla of information provided
in Senator Grassley's and my report that has ever been refuted.
It was 100 percent accurate. Yet you, as Ranking Member of the
Committee, accused me repeatedly of soliciting and
disseminating Russian disinformation. Do you want to retract
your false allegation here that you issued in your press
release on September 23rd?
Chairman Peters. No. Let us focus on what we are trying
to----
Senator Johnson. I am focusing on this because this is
exactly the type of harm we can do to our political process
when you have these big tech companies engaging in political
debate, censoring one side of the political spectrum and
amplifying the false allegations of another side. Do any of you
want to dispute that?
Mr. Boland. Senator, I think it is important that we get
the data to know. This is why the Platform Accountability and
Transparency Act is so critical to our globe and our Nation, is
that if you were able to look at the data, to understand what
had happened from content moderation, and you were able to see
the distribution, you could compare that data across the
platforms and see what sort of impact that it had.
Senator Johnson. One part of the transparency would be to
at least have people who at least used to work or work for
these platforms to at least acknowledge the highly political
nature of the individuals that work in them. Just acknowledge
it. It is obvious to everybody. Mr. Zuckerberg spent, what,
about a half a billion dollars impacting the 2020 election?
Took over the Green Bay election system, in a highly partisan
fashion. About 95 percent of the money he spent was in
Democratic strongholds in Wisconsin.
Can we at least acknowledge that there is enormous
political activity going on, partisan activity going on, within
these social media companies, rather than just trying to bury
it? Let us be honest. Let me be transparent. But let us be
honest in our transparency.
Mr. Boland. I agree with you on the request for
transparency. My experience, outside of whether someone had a
certain political leaning or not, I did not see political
leanings shape the decisions that were made inside the company,
per my experience, and what I saw.
Senator Johnson. OK. I sat it in their censure of the New
York Post article prior to the 2020 election, and I think it is
pretty obvious.
Thank you, Mr. Chairman.
Chairman Peters. Senator Lankford, you are recognized for
your questions.
OPENING STATEMENT OF SENATOR LANKFORD
Senator Lankford. Thank you, Mr. Chairman. Mr. Cain, you
have spent a lot of time studying authoritarian regimes and how
they use social media to be able to literally control their own
populations. The Chinese government obviously doing this with
the Uyghurs and what they have done. In that case you have
spent a lot of time studying and going through.
One of those features that is in TikTok, for instance, and
in several platforms, is the permissions. When you join it you
use this free platform. The user gives ByteDance, TikTok,
whoever it may be, the opportunity to be able to open their
microphone, to use facial recognition, to be able to store data
on that. How is that information used in an authoritarian
regime?
Mr. Cain. An authoritarian regime such as China will
attempt to get access to that data and use it to build
artificial intelligence capabilities, capabilities that might
involve espionage, spying on military officials, government
officials. This is a major Trojan Horse that needs to be dealt
with, and the Chinese government has made clear, in its
National Artificial Intelligence Strategies, that it does need
data, that data is its biggest target.
Senator Lankford. Right. One of the things that I have seen
from TikTok even recently is the ability to be able to keep up
with keystrokes. If you use their app to be able to then go to
other websites so they can then track your keystrokes, that
would include credit card numbers, that would include
passwords, user IDs, all of those things as well. Factual or
not factual?
Mr. Cain. Factual. You are absolutely correct, Senator.
Senator Lankford. They have made the statement publicly,
``We do not use that for any other purposes. We just maintain
that.'' That is now owned by the Chinese government. At that
point if it is going through TikTok they have access through
that to be able to get user names, passwords, facial
recognition, everything else, on this. That is the building of
a database system.
This is not some hypothetical, possible thing. This is
actually occurring.
Mr. Cain. Precisely, Senator. This is occurring. TikTok
does get gather large numbers of data, and there was one recent
study by Citizen Lab, which does work on this, which found that
TikTok does gather unusually large amounts of data from its
users.
The key login software that was found recently and reported
on, TikTok has said that they do not use this. But it is there,
and if the Chinese Communist Party wants to get access to it
they have the power to do it.
Senator Lankford. OK. Thank you.
Mr. Roetter, I want to ask you a little bit about value
system. You have a unique perspective from coming from Twitter
and then outside of it to be able to look backwards on this. It
is unusual to me that Twitter is blocked in many authoritarian
countries, but yet the leaders of those countries are allowed
to be on Twitter and to be able to put out authoritarian
propaganda, basically. They are still allowed to be able to do
that.
Twitter's value system seems to shift from country to
country, based on that country. Even if that country blocks
them from a platform, they are still allowed to be able to put
out the propaganda from that platform. Am I wrong on this? What
have you seen?
Mr. Roetter. Twitter certainly is obligated to follow the
laws of all the countries where it is operated. That is a
consequence of what you are seeing.
Senator Lankford. But it also seems to be a patchwork of
values in these countries, where in our country they will say,
``We really stand strong for this principle,'' but in another
country they do not.
Mr. Roetter. I think that is a fair characterization.
Senator Lankford. Is that a problem long-term or giving
authoritarian regimes a platform through that is just a matter
of having customers there, even if those countries actually
even block the use of Twitter in their country.
Mr. Roetter. I think the bigger problem is the consequences
of their overwhelming incentive model. Not being deeply
specific about that example, of per-country variation. That is
a consequence of trying to get everyone to use the platform and
then being subject to some constraints, whether it is local
governments or some other constraints.
I think the broader problem is the consequences of who sees
what content and what that does to people in the real world as
a function of the incentive structure that is created for these
companies.
Senator Lankford. Recent testimony on Twitter has come out
that they have had on their time Chinese spies, individuals
that work for the Indian government, individuals that work for
the Saudi government, that were on the staff and were funneling
information back to those authoritarian regimes from the staff.
I would assume Twitter has a process of actually going
through and vetting their employees. I am making the assumption
that while you were there you saw some of the vetting of how
this actually happens. How are they vetting their employees to
be able to evaluate individuals so they do not end up with
Chinese spies, Saudi employees from the Saudi government, or
required from the Indian government, for instance?
Mr. Roetter. It may have changed in the time that I was
there. There are background checks and other things you go
through when you get hired, but there was nothing I saw that
made me think that process was designed to counteract a threat
model of governments inserting spies. It was much more
pedestrian of a process than that.
Senator Lankford. It seems to be a different issue when you
talk about the Indian government saying we are requiring our
individuals to actually be on your staff, to be able to be in
the process, or to allow individuals, as was accused from the
Saudi government, to be there on the staff. That does not seem
to be a vetting issue. That seems to be a requirement. If you
are in our country we also require backdoor access, basically.
Mr. Roetter. I am not sure I am familiar with the specific
rules from India and Saudi Arabia in terms of operating in the
country.
Senator Lankford. Fair enough. But this is an issue we will
have an opportunity in the second panel to be able to talk
about this.
Mr. Boland, you have spoken out often on the algorithms
that are out there and dealing with basically how the platforms
seem to engage angry comments. The angrier that you become, the
more it helps the algorithm to be able to engage and to be able
to place this.
I have done a recommendation to Facebook for years to say
why could not the page owner, in that sense, be able to take
the comments, if people want to make comments, those comments
come to the page itself, that individual, but other individuals
cannot see it. There is basically an option that you could
create to turn off the public viewing of all the comments. If
you want to make comments to me, and we want to have dialog,
you can do that. But it prevents people that make comments on
my page from attacking other people that make comments on the
page.
Basically what Facebook has created is a place for people
to scream at each other in a lot of the political dialog, and
it is pretty hostile, and it reinforces the anger comments to
continue to be able to drive that.
What I am describing to you, of giving the user, the owner
of a page, the option to be able to make the comments between
myself and those that are making comments so they cannot attack
each other, is that technically possible, to be able to do?
Mr. Boland. That is technically possible to be able to do.
I think an important step for a lot of the product development
work is back to this transparency point of if we could have
researchers and academics involved and evaluating the different
types of scenarios here and the tradeoffs, I think that could
move us a lot faster forward.
Senator Lankford. Yes. The value system of Facebook seems
to be that we want that engagement across and that anger and
that attack to each other because that keeps people engaged.
Instead of trying to lower the temperature and saying on this
page the temperature is going to be lower, it seems to be high
temperature in as many places as possible, and the anger emoji
or anger responses seem to build in that algorithm to continue
to accelerate coming back to that page over and over again.
Mr. Boland. Yes, and I think an important point that Alex
has made is that the algorithm knows no temperature. It does
not know if something is charged or not charged. It just knows
whether it gets a result. It knows if it gets engagement.
Without there being a qualitative view over the kinds of
content, the algorithms will just chase what they are told to
chase, and they are really good at it. They are really good at
going after the metrics that they are given, and as machine
learning techniques improve you will see more and more of that.
The idea that it chases that kind of content, if that is what
gets engagement and that still gets reaction, that is what is
going to grow in the system.
Senator Lankford. OK. Mr. Chairman, thank you.
Chairman Peters. Thank you, Senator Lankford.
Senator Romney, you are recognized for your questions.
OPENING STATEMENT OF SENATOR ROMNEY
Senator Romney. Thank you, Mr. Chairman. Mr. Roetter, your
comments about the incentives of a corporation are accurate,
which is they are trying to make as much money as they can, and
to do so they are trying to get as many eyeballs as they can.
Every newspaper, every magazine, every TV show, every
broadcaster, every radio station does the same thing, which is
how can we get more eyeballs? What gets more attention?
What we are seeing with social media is not entirely
unprecedented. I was not around in the early days of broadcast,
but I presume it was Wild West initially, and then there was
threat of government intervention to tell you who could
advertise and what they could advertise and what words could be
said and how much sex and violence could be allowed.
The industry came together and said, OK, we are going to
start grading things and establishing rules. Ultimately, from
what I understand, the government also established an entity to
establish rules for broadcasting, saying what you could show at
certain hours of the day, how much sex and violence, and so
forth, could be on broadcast networks.
We have not done that with regard to social media. Social
media is far more engaging and captivating of our young people,
as well as many adults, than broadcast was. I wonder whether we
need to do this, one, whether the industry should not come
together and talk about its own decisionmaking, the rules,
where they draw the lines, and say, yes, these are things we
are all going to agree to. If the industry does not do that
whether government should, whether we should establish an
agency to say, hey, these are the rules, and you are all going
to have to follow them.
Is the industry willing to take action of that nature?
Could it? Should it? If not, should government? I will ask you,
Mr. Roetter, first, and then Mr. Boland, and Mr. Cain.
Mr. Roetter. I think probably the best predictor here is
just past behavior. My observation is the industry will share
information, which is not the information you would share if
you were generally interested in providing transparency. Brian
has talked somewhat about sharing an average, when you have a
distribution which is so non-uniform that an average is not a
useful statistic. We see that. We see exact numbers being
shared.
I view this as a two-step process. The first is we actually
do not know what is happening on these networks today, and that
is why a lot of the conversation about networks devolves into
cherry-picking examples to prove something that I believe these
networks are doing. You can also prove that they have a bias in
the other direction, depending on which example you cherry-
pick.
The first step is we need statistically representative,
unbiased, raw data that can be processed and then we understand
what they are actually doing. I think if we had a better view
of what they are actually doing in a representative way we
could then talk about do we believe their incentives are going
to create the right outcome, what is the true impact of that.
Then, faced with that shared understanding of what the networks
are doing, it is possible, to your point, that maybe the
companies would come together and decide to self-regulate
because they realize the specter of someone else regulating
them is worse. I am not sure.
But I think really the first step is demanding the
transparency so we have a shared view on what is actually
happening on these things.
Senator Romney. Yes. Thank you. Mr. Boland.
Mr. Boland. Yes, I think that usually you see industries
take self-regulatory steps when they feel like external
pressures or legislation is impending, and I do not believe
that they feel that with the United States today. I think that
drives decision from them to kind of let the status quo go.
I think what is particularly terrifying is that we do not
know what is happening on these platforms today. The nice thing
about broadcast is that you know what was broadcast. Everyone
could see what was being broadcast. With the way that our feeds
work today you have such a distribution of content that it can
look, on an average basis, that things are getting better. The
industry can tell you, ``We are making improvements. Here is
the average that shows you what we are doing.'' But for the
person who has the 99th percentile most hate-filled feed or the
group of people that have that 99th percentile most extremist
feed, they may be seeing an increase in the types of harmful
content that we will never know.
Meta provided a lot more transparency three years ago than
they provide today; that transparency is decreasing--none of
the other platforms are taking steps in that regard to increase
transparency, and you have an increased TikTok-ization of media
in that Facebook is now moving toward a TikTok model, where it
is not just friends and family content. It is unconnected,
algorithmically driven content. These kinds of fringes, these
pockets, are going to grow. We will never see them unless we
mandate that we get to see them.
Senator Romney. Thank you. Mr. Cain.
Mr. Cain. Thank you, Senator Romney, for the question.
TikTok has seen an alarming number of leaks get out in the
press, which does suggest, and I have spoken personally with
former TikTok employees who have been disgruntled, who say that
the company is not transparent, that it is not doing what it
promises. Since it has this connection to the Chinese Communist
Party it will not be transparent about what it is doing, that
this algorithm, we do not know what is behind the curtain.
China is not a place that values transparency. It is a one-
party, authoritarian state, and the most sophisticated police
state in the world. I do not think we can count on a company
such as TikTok or its parent, ByteDance, to do anything that
will actually address the problems at hand.
I think that, to be honest, this Committee is uniquely
placed to address this problem of transparency because the
subpoena power that can be used here I think would require
TikTok to open up its emails, show us what is really going on,
and show us what the China-based executives are saying with the
American executives.
Senator Romney. Yes, I must admit I share your view in that
regard, although I am probably even more alarmist than you,
which is I question whether we should allow an authoritarian
regime to have a social media capability of the scale they have
in our country, gathering the data they have. I think that is a
huge risk for us.
I have a lot of kids and grandkids, particularly grandkids
these days. I am very concerned about their exposure to social
media. Have other countries figured a better way to try and
reduce the draw and the compelling nature of social media? I
understand China, for instance, between various TikTok segments
they have a five-second gap where the screen just goes blank or
something. Are we not doing even what other nations are doing
to try and protect our kids? I will let anyone that wants to
respond to that. Maybe Brian, if you want to take that.
Mr. Boland. Yes, it is a tricky one in that you are dealing
with incentives you would mandate versus steps you would like
companies to take on their own. There are steps. You are
describing friction, right, that slowdown----
Senator Romney. Right.
Mr. Boland [continuing]. The process. There are known steps
you can take to introduce friction in the products.
Senator Romney. Have some other nations done some of those
things?
Mr. Boland. I do not know of mandated friction. I think
Europe has done a very good job of starting to move toward
required transparency, so we will see how that moves. But I
have not seen the prescriptive type of products.
Senator Romney. Thank you.
Chairman Peters. Thank you, Senator Romney.
Senator Hawley, you are recognized for your questions.
OPENING STATEMENT OF SENATOR HAWLEY
Senator Hawley. Thank you, Mr. Chair. Thanks to the
witnesses for being here.
Mr. Boland, let me start with you. Can you tell me, when
were you at Facebook?
Mr. Boland. I was there 2009 through November 2020.
Senator Hawley. November 2020. Was it normal in the time
you were at Facebook for executives or team members at
Facebook, not even have to be executives, to coordinate closely
with the United States government?
Mr. Boland. I am not aware of that.
Senator Hawley. You were never in any such meetings?
Mr. Boland. No.
Senator Hawley. You never had any contact with U.S.
Government employees in your time at Facebook?
Mr. Boland. Not that I can recall.
Senator Hawley. Would you be surprised to know that on July
16, 2021, an employee at Facebook wrote to the Department of
Health and Human Services (HHS) saying, ``I know our teams met
today to better understand the scope of what the White House
expects from us on the misinformation going forward.''
On July 23, 2021, Facebook employee thanked HHS for
``taking the time to meet earlier today. I wanted to make sure
you saw the steps we just took this past week to adjust
policies in what we are removing with respect to
misinformation.'' That included, and I am quoting, ``increasing
the strength of our demotions for Coronavirus Disease 2019
(COVID-19) and vaccine-related content.''
On April 16, 2021, Rob Flaherty at the White House
circulated a Zoom meeting invitation stating, ``White House
staff will be briefed on vaccine misinfo.''
On April 7, 2021, a Facebook employee thanked the Centers
for Disease Control and Prevention (CDC) for responding to
misinformation queries. ``We will get moving now to be able to
remove all but that one claim as soon as the announcement and
authorization happens.''
On July 28, 2022, this year, a Facebook employee reached
out to CDC about ``doing a monthly misinfo/debunking meeting.''
CDC responded, ``Yes, we would love to do that.''
On May 11, 2021, Facebook employees organized a be-on-the-
lookout meeting with CDC officials.
On July 20, 2021, Clark Humphrey at the White House emailed
Dave Sumner and others at Facebook asking, ``Any way we can get
this pulled down,'' and cited an Instagram account. Within 46
seconds, Facebook replied, ``Yep. We are on it,'' and down the
account went.
Is that normal? Is that normal in your time at Facebook?
Mr. Boland. I do not have experience around that.
Senator Hawley. You have no knowledge of anything like
this. Nothing like this ever happened, and then, presto, it
started happening just suddenly in 2020, as soon as you left?
Mr. Boland. I did not have personal experience with that,
or I did not hear about it.
Senator Hawley. You do not know anything about it at all?
You have never heard of anything like this happening, ever?
Mr. Boland. I do not.
Senator Hawley. That is remarkable. I thought that you were
the former Vice President of Partnerships Product Marketing,
Partner Engineering, Marketing, Strategic Operations and
Analytics at Facebook.
Mr. Boland. That is true.
Senator Hawley. None of this ever happened. Why did it
start happening, do you think, as soon as you left? What do you
think drove this kind of collaboration, where you have Facebook
becoming an arm of the United States government, more
specifically the White House, to censor private information,
personal speech, at the behest of government officials?
Mr. Boland. It is hard for me to comment on the specific
context of the content we are talking about, whether it was
public content or whether it was personal content. I do know
that from what I have read, and probably the same documents
that you have access to, that there were a lot of steps taken
around COVID response and COVID misinformation that may have
presented a unique scenario and a unique situation where the
company took steps to coordinate that way.
Senator Hawley. Took steps to coordinate, by which you mean
to censor the speech of ordinary Americans at the best of the
President of the United States and his Administration.
I commend to everyone who is interested these emails which
were discovered as part of litigation led by the State of
Missouri and other States as they are suing these tech
companies, including your former employer, Mr. Boland, which,
for the record, is one of the worst companies in America. They
have discovered this trove of information, extensive
coordination, extensive, between Facebook and the Biden
administration, targeting the speech of ordinary Americans. By
the way, for standards that are ever-shifting.
Early on, if you questioned that COVID had anything to do
with a lab you were marked as disinformation, you were
censored, only to have the President of the United States later
admit the possibility that COVID has some lab nexus is, in
fact, a very distinct possibility that our intelligence
communities (IC) think is actually quite a viable theory. We
have seen the same thing with people who have questions about
masks, who have questions about vaccine efficacy. It is really
quite remarkable.
Let me ask you this. What safeguards, when you were at
Facebook, were in place to protect Americans from having their
speech censored or having government censors like this access
personal information?
Mr. Boland. During my time there my experience was the
company was more reluctant to take down speech and very careful
about trying to remove content. I also do not think the company
studied the content on the platform as heavily as you would
like.
Senator Hawley. They did not do things like, not like what
they were doing later, when they were looking at particular
private Instagram accounts and removing them at the behest of
the White House? You are saying that did not happen while you
were there?
Mr. Boland. That is not a scenario that I ran across. It is
hard for me to comment on COVID pandemic response, which I
think a lot of things were outside of the norm.
Senator Hawley. I will just say this. I find it hard to
believe that suddenly Facebook became an entirely different
entity and was interfacing with the United States government in
an entirely different way only when COVID happened. I mean
maybe, but I doubt it.
Let me ask you this, Mr. Roetter? You were an engineer at
Twitter. Is that right?
Mr. Roetter. Correct, yes.
Senator Hawley. You were the Senior Vice President for
Engineering?
Mr. Roetter. Yes.
Senator Hawley. Yesterday Mr. Zatko testified to another
Committee I sit on that 4,000 engineers at Twitter had access
to all of the personal information, user data, geolocations, of
Twitter users. Is that accurate?
Mr. Roetter. I have never met him, and he joined the
company after I left, so I do not know if that particular claim
is accurate.
Senator Hawley. But he said all the engineers. You were an
engineer. Did you have access to user data?
Mr. Roetter. When I was there, I do not know if it was all
the engineers.
Senator Hawley. Did you have access to user data?
Mr. Roetter. I was head of engineering for the whole
company.
Senator Hawley. Did you have access to user data? I am
looking for a yes or a no.
I will just remind you.
Mr. Roetter. No.
Senator Hawley. You did not have access to user data?
Mr. Roetter. I think I could have gotten it.
Senator Hawley. I am sorry?
Mr. Roetter. I think I could have gotten it.
Senator Hawley. OK. If you can get it, that is what we call
access. You did have access to user data. Is that a yes?
Mr. Roetter. When I was there I probably could have. Yes,
so that is probably right. Yes, I probably could have.
Senator Hawley. OK. You did. Did you ever access any user
data?
Mr. Roetter. No.
Senator Hawley. Were you aware of Twitter engineers ever
doxing members, of users?
Mr. Roetter. No.
Senator Hawley. Were you aware of Twitter engineers ever
taking over an account and tweeting out or altering the content
of that account? Mr. Zatko said he thought that had happened.
Mr. Roetter. I am not aware of that.
Senator Hawley. OK. Lots to unpack there. Thank you, Mr.
Chair.
Chairman Peters. Thank you, Senator Hawley.
Senator Rosen, you are recognized for your questions.
OPENING STATEMENT OF SENATOR ROSEN
Senator Rosen. Thank you, Chairman Peters, Ranking Member
Portman. Thank you to the witnesses for being here today.
Transparency and accountability, I guess those are the
words of the day, because we know that social media companies,
of course, what we do. I am a former computer programmer. Data
is power. How you analyze the data, the data tells a story if
you are smart enough to listen to it.
You collect the demographic behavioral data from consumers
in order to enhance the predictive engagement algorithms to
target the consumers with ads, recommendations based on other
content, perceive interests, or even vulnerabilities. This is
really great when you are shopping for a new outfit or some new
furniture. Maybe not so great when you are on an extremist or
violent website or harmful or illegal content.
When it comes to that harmful or illegal content there has
to be greater transparency into the platform promotion
mechanisms, and how the content ultimately spreads from
platform to platform. We have small businesses, hospitals,
schools, everyone are on these platforms in some form or
fashion.
When we say ``consumer,'' we can go from the individual
right up to our full national security, that we understand
better the algorithms that amplify the content and how these
things reach their feed. Some social media platforms, for
example, they have standards in place for moving content that
promotes Holocaust denial or distortion. They are often
inconsistent with implementing the policies, but the content
flourishes.
I am going to cut right to the chase. Mr. Roetter and then
Mr. Boland, is there a difference now in how predictive user
engagement algorithms behave for harmful, illegal, or extremist
content versus other content, and how might we modify or
regulate an otherwise agnostic algorithm--It is math; it is
agnostic, to your point--an agnostic algorithm to identify
illegal, certainly illegal, or extremist content? How do we
take the agnostic out of the math? Mr. Roetter.
Mr. Roetter. Today the algorithms are doing exactly what
they are incented to do, which is maximize attention on the
platform. If you changed what those companies were accountable
for, these companies are very smart. They have a lot of
engineers and a lot of money and a lot of computational power.
They would change what the algorithms do.
For example, if companies were penalized for sharing
certain types of content, these algorithms would no longer
promote that content because it would be not optimal for them
to do so. The extra benefit they get from the attention and the
usage would be outweighed by whatever the penalty was.
This is all possible. I think the two takeaways are, one,
without transparency we are not going to know what it is doing
today, and two, they will behave optimally in the face of any
incentive structure they have. Today it is just maximizing
intention, and they are doing exactly what you would expect,
but you could change that.
Senator Rosen. Thank you.
Mr. Boland. I think it is important to note that not only
do we not know what is happening on the platforms, the
platforms do not know what is generally happening on their
platforms. The turning point for me, to go from having concerns
to being publicly vocal about my concerns, was when Facebook
said, ``Nothing on January 6th happened on our platform.'' Then
it turns out, after the fact, that we find out that there is a
lot of Stop the Steal on the platform, and there were internal
concerns around it.
In order to change these algorithms part of it is
understanding what is happening, and as a society having
conversations about what do we think the right distribution is.
Facebook has proven that with things like QAnon, after the
fact, after the fire was lit and burned through, they could
then adjust it and actually manage the distribution of that
type of content. It is possible when we know what we are
managing toward.
The problem is that it is all after the fact. It is all
after the damage has been done that you then go back and say,
OK, there has been this set of articles or conversations, and
finally we go back to address them, rather than saying we have
a whole community of researchers and people who can quickly
spot thing, raise the issues, and then adjust them. It is
doable. It is an incredibly hard problem. Like I am very
sympathetic to the fact that human speech is very complicated
and very nuanced.
Senator Rosen. But the platforms have an unwillingness.
They actually want to have this lack of understanding so they
have some deniability on the back end, if that is what you are
saying is true. We do not know it is happening. Oh my gosh, it
happened after the fact. Their lack of wanting to do the
analysis ahead of time and understand their own platform, they
are setting themselves up for deniability, in my estimation.
But we are going to move on to cybersecurity because I have
a few minutes left. We know the whistleblower complaint from
Twitter's former head of security depicted they were unable to
protect the 238 million users, government agencies, influential
figures, heads of State, from spam and security breaches. The
complaint alleged the company servers were running out-of-date
and vulnerable software and withheld dire facts about the
breaches and lack of protection for user data.
I am really concerned about cybersecurity. Companies are
laser-focused on growth, not laser-focused on protection, in my
estimation, so individuals--again, small businesses, hospitals,
schools, critical infrastructure, all of those things we are
responsible for her are at potential risk.
Again, both of your experiences working at Facebook and
Twitter, is cybersecurity a high enough priority for the large
social media platforms, and do the social media platform
security teams, do they work alongside product development,
application development to protect cyberattacks? Do you have a
hunt forward? Are you looking for these breaches? How are you
working that, and how does this threaten our own security, even
our national security?
Mr. Roetter. The teams, they do work alongside engineering,
but it is not a primary driver the way product and growth and
revenue is. You need to build something that drives usage and
revenue and then make it secure enough.
In terms of your question, is it a high enough priority,
the answer to that can only be known if you know the nature of
the threat and if the bad actors trying to break in are being
successful.
Senator Rosen. There is no hunt-forward operations built
into these things for people trying to breach the data. There
is no kind of hunt forward. There is no way that people are
really actively looking for data breaches. You are finding it
after the fact, in many cases.
Mr. Roetter. No, there are in some cases penetration
testing and people trying to simulate breaking into something
to learn. That happens.
Senator Rosen. Can you speak to that, Mr. Boland?
Mr. Boland. My sense from Meta's standpoint is that they
are quite good and quite invested in protecting people's data,
and from a cybersecurity standpoint, which goes to show you
that where there is a will and a desire to make progress on
issues I believe they can. This is an area where, in my
experience, they were quite strong.
Senator Rosen. Thank you. I see my time is up so I will
yield back. Thank you.
Chairman Peters. Thank you, Senator Rosen.
Senator Hassan, you are recognized for your questions.
OPENING STATEMENT OF SENATOR HASSAN
Senator Hassan. Thank you, Mr. Chair and Ranking Member
Portman for this hearing, and thank you to our witnesses for
being before the Committee today. I really appreciate it.
I want to start with a question that builds a little bit on
what Senator Rosen was discussing, and this is to Mr. Boland
and Mr. Roetter. Terrorists have horrifically livestreamed
their attacks on social media. These livestreamed attacks, in
turn, inspire other individuals to commit future attacks. Are
there ways for social media companies to quickly identify
terrorist content so that it is not shared in real time?
Mr. Boland, we will start with you.
Mr. Boland. I know that particular for livestreamed videos
that Meta has put considerable resources in AI to try to spot
these types of attacks and take them down quickly. I think they
have gotten a lot better, obviously, than they were with
Christchurch years ago. I think it is an incredibly hard
problem. I am not an expert on the extent of what is possible
there, but I do think they have made strides.
Senator Hassan. Mr. Roetter.
Mr. Roetter. It is certainly possible. It is a hard
technical challenge but you can build algorithms to figure out,
in real time or near real time, the content of videos. They
will not be perfect, like any sort of classification or
segmentation algorithm, but you could certainly do so.
Senator Hassan. Yes. I mean, this is an ongoing issue
because, of course, we are seeing the acceleration from idea to
action happening much more quickly in part because of the
influence of social media too. I thank you for that and I would
look forward to following up with you both on it.
Another question for the two of you. Facebook is currently
running an advertising campaign which is touting the thousands
of employees and billions of dollars that the company says it
spends on safety and security. These numbers, however, are
pretty meaningless without proper context, right?
What specific information or metrics should these companies
provide this Committee to help us fully understand their actual
commitment to safety and security? Mr. Boland, again I will
start with you.
Mr. Boland. Yes. You are 100 percent correct on the context
of the numbers matter. When they first announced their $13
billion over five years safety and security number it was in
the context of $50 billion in stock buybacks, so like a massive
imbalance of investment. They also will give you numbers of
employees. Numbers of engineers matter. If you think about
these issues, you can have employees who are non-technical who
can be in what would be a review queue or a process to look at
content. But the really important thing is engineering
resources and how many engineers are put on these problems.
I would really try to get from the companies an
understanding of where they allocate their engineers, for these
types of problems. They do not need to show you their entire
organization chart and you get to know how many are working on
Metaverse and whatnot, but these are the numbers working on
these safety and security issues, these safety issue, and this
is how they are allocated by country, by topic, et cetera. I
think that is justifiable to understand, and to feel like we
have a sense of whether that is adequate relative to the total
number of engineering employees.
Senator Hassan. OK. Thank you. Mr. Roetter.
Mr. Roetter. I think what is important is that we get
metrics that are of the form that show what results they are
getting, not metrics that basically equate to, ``We are trying
really hard. Give us a break.'' That would never work in Wall
Street. You can tell them we tried really hard to make profit
this quarter. You have to actually show what the results are.
If you have transparency over the content and how the
content is shared, and the engagements on that content, we will
be able to study. Independent people can look and see certain
content spreads very widely, other content does not--and then
after this investment that they have made has this changed or
not?
We need metrics where we can measure the actual result, not
just, ``Oh, I tried really hard, so please be happy.''
Mr. Boland. I am sorry. One more quick thought there, is
that I worry that a lot of times, because it is so painful, we
focus on these extreme examples of content, or the livestream
shootings. There is a broad swath of content that influences
people that does not feel as scary. That is the stuff that
terrifies me, and that is the stuff we do not get to see
without transparency.
Senator Hassan. OK. I thank you both. I thank the whole
panel for your testimony, and I am very grateful for this
hearing, and I will yield back.
Chairman Peters. Thank you, Senator Hassan.
During your opening statements I think each of you
discussed the product development process at these companies,
and we have talked at length about that process through the
hearing.
Mr. Boland, you discussed how Facebook does not incentivize
limiting the spread of harmful content but, of course,
prioritizes growth and revenue. Could you tell the Committee
generally what metrics inform employee compensation at the
company? What goes into that?
Mr. Boland. Employees at Facebook, it is about rewards,
right, so the rewards that you receive is your cash
compensation, your stock compensations, and promotions.
Generally if you are building products you are rewarded on the
success of that product, and that product success is defined by
some set of metrics around whether that product is being used
more.
Let us pretend that you are building a video product. The
things that you will care about are the metrics around what are
the total watch hours, how many hours are being spent watching
videos, what is the user growth, how many people are using that
video product, where is that spread geographically, et cetera.
You are incentivized on those hard metrics, and then you are
not incentivized around, what kind of content are you growing
your video with? What is the stuff underneath the hood that is
showing up, that is driving this growth? That is not your
problem. That is somebody else's problem.
The problem is that does not drive individual behavior.
Company goals are kind of there. You do not think about them.
You think about what you, individually, and your team are
goaled to deliver. That is always metrics. That is always
product growth metrics and success metrics of the product and
not success in are we keeping people safe.
Chairman Peters. There is not a trust and safety metric?
Mr. Boland. For the trust and safety team, actually my
understanding is they have been disbanded and moved into a
central team. I did not experience products like video or
others carrying a metric that was incentivizing trust or
safety.
Chairman Peters. That is not there. Mr. Roetter, is that
the case as well?
Mr. Roetter. Yes. I agree with all that. There is that
promotion system, compensation system, review system. The
problem with trust and safety metrics, typically companies may
have five top-level goals, let us say, and maybe one of them is
trust and safety. The problem is that is at odds with the other
metrics, and the other metrics always win.
If I am an engineer building, say, a new livestreaming
video service, if I launch that product and it gets some usage,
that is a feather in my cap. That is something I can say that I
did. I can point to its effect. That will help me with
promotions, compensation, career advancement.
If, at the last minute, I decide not to launch that product
because I realize I cannot control some of the safety aspects
and we should not do it if we cannot do it without certain
safeguards, I get zero credit for that. It is as if I have done
nothing for the company over the last X months.
Chairman Peters. So you are, in effect, punished, and your
future advancement will probably be questionable as well.
Mr. Roetter. A product that I build and then I do not
launch because it might not be safe is no different than if I
just did not show up to work, in terms of the future credit
that I get.
Chairman Peters. Not a good place for an employee to be.
Mr. Roetter. Correct.
Mr. Boland. You can change incentives and you can change
the way that people show up, but not even just through goals
but through process. There was an example where when Facebook
started as a desktop site and moved to mobile, Mark Zuckerberg
required that all products that were demoed to him showed
mobile in their demonstration, that they had designs around
that. He kicked the first team out. They came in without that
design, and suddenly everybody was thinking about mobile
designs.
If, in your process, you create an incentive where you
said, as part of every product design discussion is what are
all the harmful ways this product can drive hate or drive
extremism or drive polarization, you would have a radical
change in the way that people showed up to those meetings, and
in the process thought about the negative impacts of their
product.
Chairman Peters. All right. Mr. Boland, before you left it
is my understanding--and correct me if I am wrong--that you
voice objections about how Facebook recommendation algorithms
were actually promoting extreme, hateful, and racist content.
Is that correct?
Mr. Boland. That is correct. My concern was specifically
around racist content.
Chairman Peters. What was the reaction from your senior
leadership within Facebook when you expressed these concerns?
Mr. Boland. It was disappointing. I raised issues,
particularly around the distribution of racist content that I
was seeing in the CrowdTangle tool, and my concerns that we did
not understand it, and brought forward three steps that I felt
would be very good internal steps to actually help mitigate the
problem: one, more internal researchers, two, more data to
external researchers, and three, beefing up CrowdTangle to
share more information.
I had a range of responses from, ``You are wrong and that
is not the case that this is driving this,'' with no evidence,
mind you. Just, ``I believe you are wrong,'' and no counter-
evidence. Too, some, ``Yes, this might be a problem but not
something that we are working on right now.''
Chairman Peters. But when you say ``no evidence'' and those
statements that say you are wrong, you worked for a company
that looks at a lot of data and makes decisions based on data,
but this is something they wanted to ignore, basically.
Mr. Boland. Yes, so a particularly concerning moment for me
was when I had my moment where I really came to terms with
believing that the product could be causing harm. I started to
look at a variety of things that research teams were doing
internally, to understand what they were seeing.
The internal dialog and the internal documents, many of
which Frances Haugen has shared, were troubling. There was a
particular document that was an overview of polarization
research from, I think, June 2020, and that talked about
political polarization. One of the lines said, ``We have not
researched and have very little understanding of racial,
ethnic, or religious polarization.'' That underinvestment was
significantly concerning to me.
Chairman Peters. Yes. Mr. Roetter, one of the documents
submitted by a Twitter whistleblower to the U.S. Securities and
Exchange Commission (SEC) last month was a 2021 study that he
commissioned of the site integrity team's capabilities. The
study found that Twitter planned to launch a new product,
Fleets, just weeks before the 2020 elections.
The integrity team, according to that document--and I am
quoting the document--said, ``Had to beg the product team not
to launch before the election because they did not have the
resources or capabilities to take action on misinformation or
disinformation on the new product.'' The report also found,
``While product teams do elicit feedback for new product
launches, product managers are incentivized to ship products as
quickly as possible and thus are willing to accept security
risk.''
Are these findings consistent with the pattern of
decisionmaking that you saw?
Mr. Roetter. With the caveat that that specific example
happened after I was there and I cannot speak to it, that is
absolutely consistent. In fact, I would be surprised, given the
incentives at play, if the product team had done anything else.
One way we used to talk about product managers is they are,
the ``mini CEOs'' of their product, and they get consultation
from other teams--trust and safety, legal, finance,
compliance--but it is their decision to launch or not. Again,
there is no possible credit or reward for not launching,
whereas there is possibly a credit or a reward from launching.
Because they probably had more confidence that it would at
least get some usage and potentially drive revenue, there is
every reason to launch and not worry about the other issues.
Chairman Peters. Thank you.
Ranking Member Portman, any remaining questions?
Senator Portman. Let me follow up on that particular issue.
Twitter Spaces, an audio function newer to the platform, was
allegedly rolled out in such a rush, to your point, that it had
not been fully tested for safety. Twitter lacked real-time
audio content moderation capabilities when they launched it.
We are told that in the wake of our withdrawal from
Afghanistan it was exploited by the Taliban, and Taliban
supporters used this platform to discuss how cryptocurrency can
be used to fund terrorism.
First of all, is that accurate? Mr. Roetter, maybe I will
start with you. Second, is that common for Twitter to launch
products that lack content moderation capabilities? You said
that sometimes they are under pressure to ship products as soon
as possible. Was that why this happened?
Mr. Roetter. I is accurate that they are under pressure to
ship products as soon as possible, and Twitter, in particular,
has a history of being very worried about user growth and
revenue growth. It is not the runaway success that Facebook or
Google are, and so there was often very extreme pressure to
launch thing.
A saying we has is that if you walk around and ask enough
people if you can do something, eventually you will find
someone who says no. The point of that was really to emphasize
you just need to get out and do something.
Again, the overwhelming metrics are usage, and you would
never get credit or be held up as an example or promoted or get
more compensation if you did not do something because of
potential negative consequences on the safety side or
otherwise. In fact, you would be viewed probably as someone
that just says no or has a reason not to take action. There is
a huge bias toward taking action and launching things at these
companies.
Senator Portman. Yes. Are you aware of this Twitter Spaces
issue and the Taliban having exploited it?
Mr. Roetter. That specific example I am not.
Senator Portman. OK. Do you think, assuming my example is
correct, which I believe it is, that PATA would have been
helpful there, to at least get behind the curtain and figure
out why the decisions are being made?
Mr. Roetter. I have not read the draft of that, but if my
understanding is correct, yes, having more understanding of
what these products do and what sort of content is promoted and
what the internal algorithms are that drive both decisionmaking
and usage of the products, I think that would be extremely
valuable. Without any of that I would expect examples such as
this to keep happening.
Senator Portman. On this trust and safety issue, and
specifically the product development and business
decisionmaking processes--Mr. Boland, I will maybe direct this
to you--Meta disbanded its responsible innovation team just
last week it was announced. Did you see that?
Mr. Boland. I did. It was extremely disappointing.
Senator Portman. My understanding is they have been tasked
with addressing harmful effects of product and development
processes. You are saying it was concerning to you. Why are you
concerned about it, and tell us about how you interacted with
integrity teams while you were at Facebook.
Mr. Boland. I know the people who led that team. Very high
integrity, very intentional about responsible design of
products, as the team was named. Without that kind of center of
excellence that is helping to shape other teams I fear that
Meta is not going to continue to have that as a part of their
conversations.
You can think about that group as influencing and
indoctrinating, if you will, the engineers that come to the
company of how to start to think about some of these issues. It
is less hard-coded into the incentive structure, which I think
is a missing element, but would have driven really important
conversations on how to ethnically design products.
I do not believe them when they say that they are making it
a part of everything, that they are going to interweave it into
the company. That is a very convenient way to dodge the
question in my view. I do not believe that they are going to
continue to investing in it if it is not a team.
This comes at a time when Meta building the Metaverse. We
do not know how the Metaverse is going to play out. I am
extremely concerned because the paradigms we have seen in the
past, that we have started to understand, around content and
content distribution, are very different in the Metaverse. That
is an area that if I were this Committee I would spend a lot of
time really trying to understand the risks of the Metaverse. It
feels very risky to me. It feels like the next space where
there will be underinvestment, and without a team of
responsible innovation helping to guide some of that thinking
that is concerning.
Senator Portman. Again, same question to Mr. Roetter, with
regard to how to evaluate these trust and safety efforts in
general, and specifically something like the responsible
innovation team and what impact it is having. Do you think that
it would be helpful to have this legislation called the
Platform Accountability and Transparency Act?
Mr. Roetter. I think so. If we get from that more
information to illuminate what these algorithms are doing and
what the incentive structures are, that would be extremely
helpful.
I think today we are operating in a vacuum, and what we
see, a lot of the public conversation about this is people will
cherry-pick one example and use it as evidence of whatever
their theory is of these companies are doing, that, of course,
it must be true because this here is one example.
The fact of the matter is these companies are so massive
and there is so much content you can cherry-pick examples to
prove almost anything you want about these companies. Without
broad-scale representative data from which we can compute what
is being promoted and then reverse-engineer what the incentives
must be, we are never going to see a change into the things
they are optimizing for.
Senator Portman. What are your thoughts on that, Mr.
Boland?
Mr. Boland. Yes, no, I think the issue that we face today
is we have to trust, and without having robust set of data to
understand what is happening, and making these public
conversations, not company conversations is critical.
Meta would like to tell you that they do not want to put
their thumb on the scale when it comes to algorithmic
distribution. The challenge is that these algorithms were built
in a certain way that you are kind of leaning on the scale
already. You just do not realize you are leaning on it. These
algorithms today are already doing a lot to shape discourse and
to shape what people experience. We do not get to see it, and
we have to trust the companies to share with us information
that we know that they are not sharing.
As I said earlier, I think the Platform Accountability and
Transparency Act is a critical first step. We need to do it
quickly, because these things are accelerating, to understand
what is actually happening on these platforms.
Senator Portman. Mr. Cain, you have the last word.
Mr. Cain. I do believe that there are a number of issues
that were addressed here today have will have significance not
only for our democracy within America but the position of
America in the world. There has been just major changes that I
have seen personally, having been in China and Russia and
recently Ukraine, in the world of technology, in the world of
social media. My greatest concern is that we are ceding too
much ground to authoritarian regimes that seek to undermine and
malign us in whatever way they can.
The software that we are using, the AI, the apps, these are
ubiquitous. This is not the Cold War where we had hardware, we
had missiles pointed at each other. Now we have smartphones,
and it is entirely possible and quite probable that the Chinese
Communist Party has launched major incursions into our data
within America to try to undermine our liberal democracy.
Senator Portman. That is a sobering conclusion, and I do
not disagree with you, and I appreciate your testimony and our
other experts. Thank you all.
Chairman Peters. Thank you, Ranking Member Portman.
Let me just follow up on a brief question on transparency.
It is pretty clear we need transparency. We need to have,
though, the active involvement of researchers that use that
data and researchers, whether academics or civil rights groups
or rights organizations, journalists, everybody has to be
engaged.
One pushback we could get, and I would like your response,
is do you think there are ways that we can protect user data
and still provide the kind of data that is necessary for these
researchers? Is that possible?
Mr. Roetter and Mr. Boland?
Mr. Roetter. Yes, you will get that pushback, as well as a
bunch of other pushbacks, I am sure. One, it is possible. We
can obfuscate the data. We can generate random ideas that you
can hide the personally identifiable information (PII).
Second, there are examples when third-party reviewers have
access to confidential information, and because they operate in
a professional manner and are well trusted, that does not mean
it leaks out publicly. One of the examples I gave was auditors,
in the course of certifying financial statements, see a bunch
of internal financial performance that if it leaked out would
be extremely valuable to competitors. The reason we have third-
party auditors is they are allowed to balance the public's need
for information with the company's need to keep information
private.
We could do the exact same thing. There are secure
computing environments. For example, there is a bunch of health
care data in the world, which has a bunch of personally
identifiable information and very strict legislated private
requirements around that, and that is managed in a way such
that people can extract insights from the data without
violating individual privacy. We could do the same thing here.
Chairman Peters. Mr. Boland.
Mr. Boland. It is absolutely possible and doable. There are
some hard aspects to it but it can be done. I think there are
two components that I think are favorable there. One, with the
increased TikTok-ization of service more and more content is
public content, so you are really not dealing with issues
around privacy and private data.
For private data, Meta was able to solve it with their ads
measurement system. We built a system where we could connect
the ads that people saw on Facebook with the purchases that
they bought in a physical store. We were able to do that in a
privacy-safe way. If we can do it for ads, you can do it for
these other areas as well.
Chairman Peters. Great. Thank you, and I would like to
thank the three of you for your testimony here today. You
certain provided some very insightful contributions to what is
a very important conversation. We appreciate your availability
to be part of the first panel for this hearing.
This hearing is going to resume this afternoon when we will
welcome our second panel of witnesses, the chief product
officers of Meta, YouTube, TikTok, and Twitter.
The Committee will now go into recess, and then we will
reconvene at 2:30 p.m.
[Whereupon, at 11:47 a.m., the hearing was recessed, to
reconvene at 2:30 p.m. this same day.
The Committee reconvened at 2:31 p.m., in room SD-342,
Dirksen Senate Office Building, Hon. Gary Peters, Chairman of
the Committee, presiding.
OPENING STATEMENT OF CHAIRMAN PETERS\1\
Chairman Peters. This morning the Committee heard testimony
from experts and former executives at Facebook and Twitter that
provided important transparency and context for how many of the
largest social media companies operate. Independent and
accurate information about how companies balance competing
priorities or how they do not, who within the companies make
those decisions, and how they build their products is
incredibly difficult to find.
---------------------------------------------------------------------------
\1\ The prepared statement of Senator Peters appears in the
Appendix on page 97.
---------------------------------------------------------------------------
This morning's testimony shed some light on many of the
areas that this Committee and the public have questions about.
I look forward to building on that testimony with our second
panel of witnesses, who can speak directly to what steps Meta,
YouTube, TikTok, and Twitter are taking to stop the spread of
extremist content on their platforms, and I want to sincerely
thank each and every one of you for being here today before us.
As we heard from our panel this morning, as Chief Product
and Operating officers, you play key roles in your company's
decisionmaking process. You set the agendas for the product
teams who are constantly updating the apps and developing new
features. You play a prominent role in setting priorities and
determining what tradeoffs to make among those priorities, as
product teams launch new features or make changes to the apps.
This is the first time executives in your positions have
appeared before Congress, and I really do appreciate you
joining us for this opportunity to hear directly about your
role at these very powerful companies.
The platforms you are representing today reach billions of
people around the world. Meta's platforms reach more than 3.6
billion people a month, TikTok has more than one billion users
a month, YouTube reaches almost two billion people a month, and
Twitter has more than 200 million monthly users. The reach is
massive and so is the influence that your platforms wield.
Whether users are fully aware of it or not, the content
they see on your platforms shapes their reality, and the
business decisions you make are one of the main driving forces
of that phenomena. This amount of influence may have a minimal
impact on the average user of your platform, but we have seen
firsthand how quickly dangerous and extremist content can
proliferate online, especially to vulnerable communities or
users already on the fringe, and alter how people view the
world.
Conspiracies like QAnon and Stop the Steal, hateful
ideologies like white supremacy and anti-Semitism, and so many
more examples of harmful content pollute your platforms. This
extremist content can spread like wildfire, amplified by the
recommendation algorithms and other tools your team build to
increase your companies' audiences and profits. Extremists use
the products you designed to recruit and to radicalize
followers, and plot attacks, including the January 6th attack
on the Capitol, our democracy, and our Nation.
There is no question that there is a relationship between
social media amplification of this extremist content and the
rise we have seen in hate crimes and domestic terrorist attacks
that mark one of the gravest threats to our homeland security.
Despite these serious threats, I am concerned that your
companies have still not taken the necessary steps to limit the
spread of the hateful, dangerous, and extremist content that
has motivated real-world violence.
That we all understand exactly the type of extremist
content we are discussing today and how challenging this
problem is to tackle--it is clearly a challenge--I would like
to take a moment to show a few examples, if you would check the
screen.
[Video plays.]
This morning we heard from former executives that your
companies have no incentive to effectively address the problem
this content creates or prioritize the safety of your users as
you build and introduce new social media products. Instead,
like any for-profit company, your incentives are to prioritize
user engagement, grow your platforms, and generate revenue.
I have asked you to appear before the Committee today to
answer questions about your companies' incentives and
priorities, how those incentives are reflected in how you
compensate and promote your product development engineers,
managers, and other employees, and to provide important insight
into your decisionmaking processes.
I want to thank you again for joining us today. I am
looking forward to this conversation, so that our Committee and
the public can better understand this serious problem and how
it threatens the safety and security of our Nation.
Ranking Member Portman, you are recognized for your opening
comment.
OPENING STATEMENT OF SENATOR PORTMAN\1\
Senator Portman. Thank you, Mr. Chairman. We had a very
productive hearing this morning with experts on the impact of
social media on homeland security, and I look forward to our
discussion this afternoon, and I want to thank the
representatives here from Meta, YouTube, TikTok, and Twitter.
Thank you all for being here, and in anticipation of another
good hearing I appreciate you being very frank with us today
and providing us information we need to be able to move
forward.
---------------------------------------------------------------------------
\1\ The prepared statement of Senator Portman appears in the
Appendix on page 101.
---------------------------------------------------------------------------
About 300 million Americans now use social media. We know
that social media has offered unprecedented connectivity, and
that is often very positive, but we also know it has raised
serious concerns for our children, our civic culture, and our
national security. Terrorists and violent extremists, drug
cartels, criminals, authoritarian regimes, and other dangerous
forces have used social media in furtherance of their goals.
They have exploited your platforms.
Perhaps the most concerning consequence of social media is
the ability for our adversaries to exploit platforms to harm
Americans for their own geopolitical gain. As an example, in
this second panel I hope we will discuss China's influence over
TikTok, which is a social media app that at least one-third of
Americans use, and a lot of young people.
As the lead Republican and former Chairman of the Permanent
Subcommittee on Investigations (PSI), I have been focused on
China's malign activities for many years, and in 2019 I led
year-long bipartisan investigation which found that China
recruits U.S.-based researchers to steal taxpayer-funded
intellectual property and research for its own military and
economic gain.
Following this report, I introduced bipartisan legislation,
Safeguarding American Innovation Act, which seeks to stop U.S.
taxpayer-funded research and Internet Protocol (IP) from
falling into the hands of the Communist Party of China (CCP).
Two months ago I issued a new report detailing China's
efforts to target influence and undermine the United States and
Federal Reserve. China has a pattern of economic and cyber
espionage, and social media for them is just another
opportunity. I am highly concerned about TikTok and how China
may be leveraging their influence to access the platform's data
on Americans.
Chinese law requires all companies operating under its
jurisdiction to, in essence, allow the Chinese Communist Party
to access every piece of data collected. Any company that
refuses to comply with the CCP's demand is subject to severe
consequences, as are individuals. Therefore, since both TikTok
and its parent company, ByteDance, have a presence in Mainland
China, an expert witness this morning told that TikTok's user
data could be accessed by the Chinese Communist Party. We want
to talk more about that today.
This means that the CCP may have access to 100 million
Americans' personal and proprietary information. As the U.S.
Government has warned, China's access to user data will allow
it to extend its malign agenda and build dossier's on American
citizens. The overwhelming popularity of this app with
America's youth will allow China to collect never-before
accessed troves of data on our children, the future generations
of Americans.
But the challenges that social media poses to our children
are not limited to TikTok. We continue to see the proliferation
of child sexual abuse material online. I have been at the
forefront of this for years. I am proud that the Stop Enabling
Sex Traffickers Act was signed into law in 2018. This was the
first bill to reform Section 230, by removing barriers to both
criminal prosecution and civil suits against websites that
knowingly facilitate online sex trafficking.
Because of this change in law, courts are beginning to
affirm that Section 230 cannot shield internet companies when
they fail to respond to images of child exploitation and
continue to profit from exploitation on their platforms. A
specific case against Twitter is now being considered by the
Ninth Circuit Court of Appeals, for example, and will show if
the law needs to be expanded in order to properly protect
children.
But it is not just Twitter. The fight continues on other
platforms that are used to exploit children. Meta announced
earlier this year that they would not report all explicit
images of children and would instead, and I quote, ``err on the
side of an adult,'' end quote, when moderating explicit images
of could-be children. In other words, when the age of an
individual in a sexual image is uncertain, content moderators
are told to put their thumbs on the scale of that individual
being an adult.
To me this is shocking. Let us be clear what we are talking
about. This is child sexual abuse material, images of a minor's
rape, exploitation. Somehow, at least what we have been told,
is that Meta has decided that these should not be referred to
law enforcement. The National Center for Missing and Exploited
Children (NCMEC) has made it clear that images must be reported
if they appear to involve a child so that law enforcement can
intervene and stop the abuse and prosecute perpetrators.
I worked with colleague across the aisle to draft this
legislation, Stop Enabling Sex Traffickers Act (SESTA), and we
crafted it narrowly so that it would be focused on ending
trafficking and exploitation online. But it may, in fact, be
too narrow if companies continue to turn away from keeping the
exploitation of children off of their platforms. I hope my
colleagues will take up the challenge of revisiting SESTA and
tightening the standard so that entities showing a reckless
disregard for the sexual exploitation of children are held
accountable. I am ready to be an ally in this fight, even after
I leave the Senate this term.
I look forward to discussing these matters, especially
regarding how product development processes appear to be at
odds with user safety as well as the need for more detailed
transparency from companies, and again, I look forward to the
testimony.
Thank you, Mr. Chairman.
Chairman Peters. Thank you, Ranking Member Portman.
Our first witnesses is Chris Cox, Chief Product Officer at
Meta. Mr. Cox joined Meta in 2004, as a software engineer, and
has helped build the first versions of signature Facebook
features, including the News Feed. He later served as Director
of Human Resources (HR), leading the direction and tone of
Meta's company culture.
In 2008, he began serving as Vice President of Product, and
in this role Mr. Cox built out the initial product management
and design teams before being promoted to Chief Product Officer
in 2014, and began his role overseeing the family of apps in
2016.
Before I ask you to have your opening comments, we skipped
over an important part of the Committee, and that is that it is
the practice of this Committee to swear in witnesses. If the
four of you would please stand up and raise your right hands.
Do you swear that the testimony that you will give before
this Committee will be the truth, the whole truth, and nothing
but the truth, so help you, God?
Mr. Cox. I do.
Mr. Mohan. I do.
Ms. Pappas. I do.
Mr. Sullivan. Yes.
Chairman Peters. All four answered in the affirmative.
Thank you. You may be seated.
Mr. Cox, I have already had your introductions so please
proceed with your opening comments.
TESTIMONY OF CHRIS COX,\1\ CHIEF PRODUCT OFFICER, META
Mr. Cox. Thank you, Chairman Peters, Ranking Member
Portman, distinguished Members of the Committee. Thank you for
the opportunity to appear here before you today. My name is
Chris Cox. I am Meta's Chief Product Officer, overseeing our
apps and privacy teams.
---------------------------------------------------------------------------
\1\ The prepared statement of Mr. Cox appears in the Appendix on
page 125.
---------------------------------------------------------------------------
I first joined the company in 2005, as one of our first 15
software engineers. I care deeply about the work we do to help
people connect with things and the people they care the most
about. It is important to us that we help people feel safe on
our apps, and we stand firmly against the exploitation of
social media by those committed to inciting violence and hate.
That is why we prohibit hate speech, terrorism, and other
harmful content.
To enforce these rules, we employ tens of thousands of
people and we use industry-leading technology. We regularly
publish transparency reports so people can see how we are doing
over time and how we compare to other internet platforms.
I am proud that we have invested around $5 billion last
year alone and have over 40,000 people working on safety and
security, more than any other tech company, even adjusted for
scale. Our efforts are making a difference. For example, we
reduced by more than half the amount of hate speech people see
on Facebook over the last 18 months.
People often talk about these types of issues as safety
issues, but at Meta we also refer to them as integrity issues.
Integrity is our way of referring to the work we do to prevent
bad actors from abusing our platforms. This includes working to
stop terrorists and violent extremists and also bullying and
harassment, scams, and other types of harm.
As the Chief Product Officer I am proud that safety and
integrity are key to the product experience. We build products
and continually update them with safety and integrity in mind.
It is a core part of our ethos, that as we develop products we
constantly think about how people are going to use them and
work to make sure they can do so safely.
I know you have questions about our algorithms. Like most
platforms, Facebook and Instagram use different algorithms for
various features. For example, we use algorithms to help keep
our community safe by identifying and removing content that
violates our policies, including hate speech, incitement, and
terrorism. This work often happens before anyone reports
content to us, sometimes even at the point of creation. We use
algorithms to rank the content that appears in people's feed
and search results, to help deliver relevant advertising, and a
whole lot more.
I also want to stress that our goal is to help people see
what they find most valuable. It is not to keep people on the
service for a particular length of time, and it is certainly
not to give people the most provocative or enraging content. In
fact, key parts of those systems are designed to do just the
opposite. We reduce the distribution of many types of content,
including because they may be misleading or are found to be
false by independent fact-checking partners.
At the end of the day, our job is to build the best product
for people, and that is a product that is reliable, fast, safe,
secure, and relevant, a product that connects people to content
relevant to their interests and connects them to their family
and friends. That is the product that people want, and that is
the product we wake up every day trying to build.
We appreciate your attention to these important issues and
look forward to continuing to work with you to find ways we can
continue to improve our products, our processes, and our
partnerships.
Thank you, and I look forward to your questions
Chairman Peters. Thank you, Mr. Cox.
Our next witness is Neal Mohan, Chief Product Officer at
YouTube. In his role, Mr. Mohan is responsible for YouTube
products and user experience on all platforms and devices
globally, including YouTube's core mobile applications,
YouTube, YouTube Kids and Music, YouTube Red, and YouTube TV,
as well as other designs, policies, and services.
Previously Mr. Mohan was Senior Vice President of Display
and Video Ads at Google, and prior to joining Google Mr. Mohan
served as Senior Vice President of Strategy and Product
Development at DoubleClick, an advertisement company that
developed and provided internet and ad-serving services before
its acquisition by Google. In that role, he built the company's
strategic plan, led the product management team, and grew the
business rapidly.
Mr. Mohan, welcome to the Committee. You may proceed with
your opening remarks.
TESTIMONY OF NEAL MOHAN,\1\ CHIEF PRODUCT OFFICER, YOUTUBE
Mr. Mohan. Thank you, Chairman Peters, Ranking Member
Portman, and distinguished Members of the Committee. Thank you
for the opportunity to appear before you here today. As the
Chairman mentioned, my name is Neal Mohan, and I am the Chief
Product Officer of YouTube. In my role I am responsible for all
of YouTube's products, our user experience, and trust and
safety globally.
---------------------------------------------------------------------------
\1\ The prepared statement of Mr. Mohan appears in the Appendix on
page 129.
---------------------------------------------------------------------------
YouTube's mission is to give everyone a voice and show them
the world. Our openness is core to that mission and enables us
to help billions of people around the world to learn new
skills, discover emerging music and artists, and enjoy videos
from their favorite creators.
We are also proud to be a place where creative
entrepreneurs can build thriving small businesses. Last year,
YouTube's creative ecosystem contributed over $25 billion to
the U.S. gross domestic product (GDP), and we supported more
than the full-time equivalent of 425,000 jobs across the
country.
Our commitment to openness works hand-in-hand with our
responsibility to protect our community from harmful content.
Responsibility is central to every product and policy decision
we make, and is our No. 1 priority.
To that end, I want to make clear that there is no place on
YouTube for violent extremist content. Our policies prohibit
content that promotes terrorism, violence, extremism, and hate
speech. Not only is the type of content harmful to our
community, the overwhelming majority of creators, viewers, and
advertisers do not want to be associated with it, meaning it is
also bad for our business.
In my testimony today I will provide more information on
our approach to responsibility as well as our policies and
technology that enable our skilled enforcement efforts to
combat terrorist content online.
We have four pillars of responsibility. We call them the
Four R's: we Remove content that violates our policies as
quickly as possible, we Raise up authoritative sources, we
Reduce the spread of content that does not violate our policies
but brushes up against our lines, and we Reward trusted
creators and artists. My written submission explains each of
these Four R's in much more detail.
This framework enables us to uphold our responsibility to
the YouTube community and society while preserving the
opportunities of an open platform. For us, safety and growth
are intertwined. Violative content undermines user trust and
satisfaction, deters advertisers from investing in ads on
YouTube, and harms the creators that have built businesses on
our service.
Let me discuss YouTube's policies prohibiting terrorist,
violent, and extremist content. Our Community Guidelines set
the rules of the road for content on YouTube. These policies
explicitly prohibit terrorist organizations using our services
for any purpose, and we routinely remove such material. We rely
on a combination of people and technology to enforce these
policies. In fact, machine learning is a critical tool in our
effort to remove violative content at scale before it is widely
viewed.
As a result of our ongoing investments in teams and
technology, in the first six months of 2022 we removed close to
8.4 million videos for violating our policies. Over 90 percent
of this violative content was first detected by machines, the
majority of it removed before receiving ten views.
Our policies are complemented by our work to raise up
authoritative content and reduce the spread of content that
comes close to but does not quite violate our policies. For
news and information topics our systems elevate authoritative
sources such as news outlets and public health authorities and
search results and Watch Next panels, and what we call
borderline content is not widely recommended.
We also share best practices on counterterrorism with our
industry peers through the Global Internet Forum to Counter
Terrorism (GIFCT), which is dedicated to disrupting terrorist
abuse of digital platforms. Responsibility is and will continue
to be YouTube's No. 1 priority. Our business literally depends
on it.
Thank you, Mr. Chairman, for convening this important
hearing. We look forward to continuing to work with you to
address these challenges. Thank you.
Chairman Peters. Thank you, Mr. Mohan.
Our next witness today is Vanessa Pappas, Chief Operating
Officer at TikTok. In her role she is responsible for
overseeing content, operations, marketing, and product teams.
She previously served as interim head of TikTok globally. She
also has experience serving as Global Head of Creative Insights
at YouTube, where she oversaw YouTube's global creative
research and trends, audience development, creative strategy,
and growth teams.
Before joining YouTube, Ms. Pappas served as Vice President
of Programming and Audience Development at nextnewnetwork,
later acquired by YouTube and Google, where she spearheaded
business partnerships and audience development efforts.
Ms. Pappas, welcome to the Committee. You may proceed with
your opening comments.
TESTIMONY OF VANESSA PAPPAS,\1\ CHIEF OPERATING OFFICER, TIKTOK
Ms. Pappas. Great. Thank you for having me.
---------------------------------------------------------------------------
\1\ The prepared statement of Ms. Pappas appears in the Appendix on
page 136.
---------------------------------------------------------------------------
Chairman Peters, Ranking Member Portman, and Members of the
Committee, thank you for the opportunity to appear before you
today to discuss how TikTok is delivering on our commitment to
provide a safe and welcoming experience for our community while
also combating some of the real-world harms that are the focus
of this Committee's important work.
My name is Vanessa Pappas, and I am the Chief Operating
Officer for TikTok. I live in Los Angeles with my family and I
have been in the United States for 20 years, and have spent my
career in entertainment and media. Prior to joining TikTok, I
was an executive at YouTube. I am passionate about creating
safe online communities where people can express themselves
creatively and discover entertaining and useful content.
At TikTok our focus on safety starts at the top, with a
leadership team goal to strengthen safety and build trust, and
this focus on safety and security flows through our product
decisions.
As a person who believes in the potential of online
platforms to create amazing opportunities for individuals, for
businesses, and for society, I am personally invested in this
goal, and as an executive there is no responsibility greater
than protecting the people of our platform.
TikTok's mission is to inspire creativity and bring joy,
and more than 1 billion people around the globe enjoy the
authentic, entertaining content that TikTok is known for.
We know that with success and growth comes responsibility.
We are committed to being an industry leader in safety and
security, and earning trust through the transparency of our
actions. Let me talk first about safety and security.
At TikTok, creating a safe environment means we make
decisions that prioritize the well-being of our community and
limit the potential of online polarization or real-world harm,
even if those choices come at the expense of short-term
commercial success. Our trust and safety teams have an active
seat at the product development roadmap and before launch.
Our terms of service and community guidelines are built to
help ensure our vision of a safe and authentic experiences. Our
policies have zero tolerance for disinformation, violent
extremism, and hateful behavior. Enforcement of these policies
in the United States is led by our U.S. safety team in Los
Angeles, which reports directly to me. TikTok has thousands of
people working across safety, privacy, and security, and we
invest heavily in technology to detect potential violations or
suspicious accounts at scale.
We also work to prevent the spread of harmful content. For
instance, with the help of partners, including the U.S.
intelligence agencies, we identify groups and individuals in
the United States and abroad who promote violent extremism and
hateful ideologies, and we work to eliminate that content
associated with them. Examples include foreign terrorist
organizations, drug cartels, and groups such as Three
Percenters and Oath Keepers. Anyone who searches for this
content or related hashtags or keywords will instead be
redirected to our community guidelines.
Notably, TikTok was not the platform of choice for those
who organized the violence at the Capitol on January 6, 2021.
Of 686 references in the Department of Justice (DOJ) charging
documents, TikTok was mentioned in only 18.
I will next talk about trust and transparency. Trust is a
huge component of safety, and it is hard to earn but easy to
lose. We hold ourselves to a high standard when it comes to
being transparent about our work on safety and security in
order to build trust.
You may be familiar with Project Texas, a critical and
industry-leading initiative we have been pursuing for over a
year. We are making progress toward a final agreement with the
U.S. Government to further safeguard U.S. user data and fully
address U.S. national security interests. We look forward to
finalizing this arrangement and sharing more when we are able.
Since 2019, we have released community guideline
enforcement reports which detail the type and volume of the
content we remove. For instance, in the first quarter of 2022,
more than 95 percent of the time we discovered and removed
problematic content before receiving any reports. We also
disclose the data on the requests we receive from law
enforcement and/or governments.
Finally, our Transparency and Accountability Centers show
how we moderate content and recommend content. We would be
happy to arrange for a tour for Members and Committee staff as
we have for others at your convenience. Last month, we
confirmed that our content moderation and recommendation models
will be vetted and validated by Oracle. We recognize your
questions and concerns and strive to lead the industry in
meaningful transparency.
Thank you again for inviting me today. We know that our
work in safety and security is never done. These issues are of
the utmost importance to TikTok, to our community, and to our
industry. We are glad to be a part of a forward-looking
conversation such as this one so that we can better work
together to address these critical challenges.
I look forward to answering your questions. Thank you.
Chairman Peters. Thank you, Ms. Pappas.
Our final witness today is Jay Sullivan, General Manager of
Bluebird, Twitter's Consumer Products. He concurrently serves
as the interim General Manager of Goldbird, Twitter's Revenue
Products, and previously served as Vice President of Consumer
Product at Twitter.
Prior to joining Twitter, Mr. Sullivan worked at Facebook
where he led the development of Realty Labs' AI Assistant, and
then led the privacy, integrity, and systems product teams for
Messenger and Instagram Direct, launching many user-focused
features and improvements.
Mr. Sullivan, welcome to the Committee. You may proceed
with your opening comments.
TESTIMONY OF JAY SULLIVAN,\1\ GENERAL MANAGER OF BLUEBIRD,
TWITTER
Mr. Sullivan. Chairman Peters, Ranking Member Portman, and
distinguished Members of the Committee, thank you for the
opportunity to speak with you today about this important issue.
My name is Jay Sullivan. I joined Twitter in November 2021. In
April of this year I became General Manager (GM) of Twitter's
Consumer Product team. This team is responsible for the main
features that people use on Twitter's mobile apps and website.
I am also the General Manager of Twitter's Revenue Products
team.
---------------------------------------------------------------------------
\1\ The prepared statement of Mr. Sullivan appears in the Appendix
on page 154.
---------------------------------------------------------------------------
Twitter's purpose as a company is to serve the public
conversation. The open nature of our service gives a voice to a
world of diverse people, perspective, ideas, and information.
We believe that Twitter is a force for good in the world.
In the past year we have seen people come to Twitter to get
on-the-ground information about the conflict in Ukraine, to
access lifesaving information during natural disasters, and to
exchange ideas about diverse topics ranging from news to
culture to sports. The goal of the Consumer Product team is to
increase healthy participation in the public conversation. We
measure our success by how many people use Twitter and the
health and safety of the platform. These two priorities go hand
in hand. If people do not feel protected from hate, abuse, and
harassment they will simply leave the service.
By the same token, advertisers do not want their brands,
products, or services to appear anywhere near harmful content.
They will simply pull their ads. This is why it is not in our
interest to have harmful content on our platform, and this is
why we set out to build features that promote a healthy
conversation, and why we will pause, delay, or stop a product
rollout if we have health or safety concerns.
My written testimony outlines many steps that Twitter has
taken to safeguard our service and improve health on our
platform. I would like to use this time to explain my team's
overall approach to health, which is built on three main
pillars.
The first pillar is integrating health and safety
considerations into the product design and development process.
We proactively and methodically assess risk and potential
unintended consequences before we begin development of a new
feature and through the development process.
We also develop new features that incentivize healthy
discourse. Some recent examples are the development of prompts
that encourage people to read articles before sharing them;
interstitial labels that provide context; Birdwatch, a
community-powered annotation feature; Twitter Circle; and many
more health and safety features.
But we cannot always prevent bad behavior so we also build
tools that enable us to identify and take action on harmful
content, including machine learning software to help detect it.
These tools enhance the customer experience by decreasing the
burden on individuals to report content for review, and they
improve the platform overall.
The second pillar is developing and enforcing policies
designed around health and safety. We have a team responsible
for developing the Twitter Rules, the policies and governance
frameworks that prevent and mitigate harm to the people who use
Twitter. This team does not report to me but we work closely
together.
Twitter's policies prohibit terrorist and other violent
organizations on our platform, inciting violence, harassment
targeted at individuals or groups, and hateful conduct. Our
platform integrity and authenticity policies address efforts to
spread misinformation relating to civic integrity, moments of
crisis, COVID, and synthetic and manipulated media.
The third pillar is transparency and accountability. We
directly engage with outside experts, formal public feedback
processes, and research. For example, in 2018, we were the
first in the industry to release an archive of potential
foreign influence operations identified on Twitter, enabling a
host of research on this important issue. This effort has now
evolved into the Twitter Moderation Research Consortium. We
also provide industry-leading research access to the Twitter
application programming interface (API).
Underpinning all of this work are our culture, our
processes, and our technology. I, and other senior leaders at
the company, work to encourage and empower every employee to
contribute to this shared goal.
I look forward to the discussion today. Thank you.
Chairman Peters. Thank you, Mr. Sullivan.
Mr. Cox and Mr. Mohan, the question is going to be directed
to you. The dangerous QAnon conspiracy that started in 2017,
spread unchecked on your platforms for years before you started
to downrank and then ban them on your platforms. A February
2022 poll found that 16 percent of Americans now believe in
this conspiracy theory.
We heard this morning that your algorithms push sensational
content. Mark Zuckerberg, in fact, said in 2018, that this is,
quote, ``basic incentive problem,'' end of quote, and he goes
on to say that, ``when left unchecked people will engage
disproportionately with more sensationalist and provocative
content,'' from Mr. Zuckerberg.
My question is, if your recommendation algorithms are
focusing on engagement--we understand the business reasons for
doing that--and provocative content that increases that
engagement--because that is certainly the most engaging--is it
inevitable that they will promote extreme content that you have
not yet labeled violative, like QAnon?
Mr. Cox, we will start with you, and then Mr. Mohan.
Mr. Cox. Thank you, Senator. To start with QAnon, this is
an organization that today is labeled as a violence-inducing
network and so is not allowed on our platform. In general, we
believe there is no place for terrorism, for violence-inducing
content, for extremism across the network. We work hard to take
that content down, and as we have talked about, we publish our
results and work with law enforcement to make sure that we can
do so quickly.
Chairman Peters. Mr. Mohan.
Mr. Mohan. Senator, thank you for the question. QAnon on
YouTube is deemed to be a harmful, criminal conspiracy. We do
not allow that content on our platform. We have been removing
that type of content from our platform for years, given the
nature of potential incitement to violence.
But we do not just stop there. Not only do we remove the
content because there is no place for hate, harassment, violent
extremism, or graphic violence of any kind on YouTube, but we
also make sure that when users are looking for information on
our platform around a particular topic, news topic, what have
you, we raise up content that comes from authoritative sources,
that includes typically mainstream media outlets, et cetera,
that can put that particular news story in context.
We have a combination of a number of tools, those Four R's
that I alluded to before, that work comprehensively to make
sure that this type of content has no home on YouTube.
Chairman Peters. I appreciate that and I wanted to hear
your response from both of you. I appreciate it has no home and
you are aware of that now. The intro to the question was that
it started to spread in 2017. This stuff was on your platforms
for years. It took you a long time to come to the conclusion
that both of you have just come to.
But I want to get back to, really, the question that I
have, is that we have a quote from Mr. Zuckerberg. We
understand the business model, although I think Mr. Cox said
that it is not necessarily to keep people on platforms. It is
to engage. The more people that are engaged in your platforms,
the better from a business perspective. You will be able to
serve up more ads for folks, generate more revenue. As Mr.
Zuckerberg said, people engage disproportionately in
sensational and provocative content.
That content is actually good for your business. If more
sensational, provocative content is put forward, get people to
stay on the platform longer, you are going to be able to show
them more ads.
My question is, is that not inevitable that that is going
to happen when you continue to put this content? I appreciate
after the fact, and we are going to talk about your model up
front to try to prevent some of this stuff from happening at
the front end. I appreciate at the back end that you are going
to take some action. How many people saw the QAnon false or the
conspiracy theory? I said 16 percent of the American people now
think this conspiracy theory is real. You caught it, but not
until 16 percent of the American people are part of this
insidious theory.
Tell me about the up front. Why are you not engaged up
front before you launch products, to understand and perhaps
anticipate how things could be misused?
I will start with you, Mr. Mohan, this time, and then Mr.
Cox.
Mr. Mohan. Yes, Senator. I appreciate the opportunity to
clarify what I think might be a potential misconception in
terms of our business model and what our incentives are.
To be very clear, we have no incentive to post this
content, to promote it in any nature, and that goes to the
fundamental nature of what YouTube's business actually is. We
are fundamentally an advertising platform. We generate revenue
through advertising partners. We share that revenue, the
majority of that revenue, with our creators.
When we talk about the creator economy, all those over
400,000 jobs that we have created in this country, it is
through that business model that generates revenue on behalf of
our creators. Our advertisers have told us, in no uncertain
terms, that they do not want to be associated with content that
promotes hate, violent extremism of any sort, terrorism, or
what have you. I have firsthand experience myself, over the
years. When they feel that sort of content is on our platform,
they walk away.
We have not just a moral imperative--that is my top
priority, living up to our responsibility--but also it aligns
with our business goals.
Chairman Peters. I appreciate that, and my question is how
long it takes to be able to identify. This was up for two-plus
years before the changes there.
Mr. Cox, Facebook currently has an advertising campaign
touting that you have spent $16 billion over the last six years
on safety and security. I do not think it is the amount of
money that is spent. It is about the results that are most
important. You are a very large company, and, in fact, I think
over the last six years a revenue close to $450 billion, a
massive amount of money, a relatively small amount relative to
the total revenues of your company. In fact, I think it is
equal to basically what you have spent is $1 per user per year
for the entire globe. In comparison, Meta spent over $85
billion on stock buybacks over the last six years, considerably
more than you pay on trust and safety.
Why is your company willing to spend so much more per year
to drive up a stock price but not willing to spend the money
necessary to be able to pull down this dangerous content a
whole lot quicker and perhaps actually be forward-leaning and
design products from the get-go to eliminate the abuse of these
platforms?
Mr. Cox. Thank you, Senator. This is an issue for the whole
company, not just for the safety teams or for the specific
investments that we talked about there. I expect every
engineer, product manager, designer, researcher, data
scientist, whether it is building a new product or whether it
iterating upon an existing product, to pay attention to safety.
That is something that is built into the deoxyribonucleic acid
(DNA) of the company. It is something I personally care very
deeply about. It is something that we expect folks who are
designing products to think about as a part of their work.
In addition to the specific investment of 40,000 folks who
work directly on safety and security for the company, it is
part of the culture of how everybody at the company thinks
about their work.
Chairman Peters. I am going to turn it over to the Ranking
Member because we have a number of Members here, but I am going
to drill down a little further on that comment in a further
round.
Ranking Member Portman, you are recognized for your
questions.
Senator Portman. Thank you, Mr. Chairman. I look forward to
getting into this issue of the balance between free speech and
the hate speech that leads to violence because it is a line
that has to be drawn. I know it is not easy, but I am going to
talk about one that I think is easier, and that is child sexual
exploitation. I talked about it my opening statement a little
bit.
We all know that the threat of this sexual abuse material
is a persistent threat. In fact, we note that last year over 29
million reports came in of child sexual exploitation. That was
a 35 percent increase from just 2020, so this is an increasing
problem across the board, but particularly with regard to our
kids.
That is why I thought it was so unfortunate, Mr. Cox, when
I learned about the Meta policy directing content moderators,
and I quoted this earlier, but it is to ``err on the side of
the person involved in sexual exploitation being an adult''
when they are unsure about the age of the person. Let me give
you a chance to respond to that. This has been in the public
media. It does not mean that it is true, I suppose, but is that
truly what you directed your content moderators to do?
Mr. Cox. Senator, I know that, first of all, this is an
incredibly serious issue and I appreciate your work on this
issue. As the father of two kids, this is something I
personally care about, making sure that we pay attention to as
well.
The work that we do here is in consultation with NCMEC. We
have been the most aggressive of the tech companies there. We
have referred more content to them, I believe, than all the
other tech platforms combined. That is both through the work we
do on WhatsApp and Messenger as well as across the family of
apps.
My understanding on this specific question is that we
received direction from NCMEC to prioritize known Cyber
Security Assessment & Management (CSAM) content, which was the
nudge that they gave us and where they wanted us to focus our
time. I have not been focused on that specific conversation and
I would be happy to have the team follow up.
Senator Portman. Yes. Let me make sure I understand this.
You are blaming the National Center for Missing and Exploited
Children for changing your approach of moderators saying that
we are going to assume that kid are adults if we do not know?
NCMEC has said you have a responsibility, all of you do, to
report all images that appear to involve a child so that law
enforcement can intervene to stop the abuse and prosecute the
perpetrators, period. I cannot believe that you are saying that
NCMEC would want you guys to send out instructions to your
moderators saying err on the side of this being an adult if you
are not sure.
Did I misunderstand what you said?
Mr. Cox. Senator, I have not been in that specific
conversation with NCMEC, but I would be happy to follow up on
the details. I agree it is a very important issue.
Senator Portman. Given your role, would you commit to, one,
getting back to me on it, and two, ensuring that if that is
true that you change that policy?
Mr. Cox. Senator, I could commit to getting into the
details of the policy and make sure we follow up with the team
to work on it.
Senator Portman. OK. You are the Chief Product Officer. I
would hope that this is one that you would follow up on and
ensure it is not the direction you are giving your moderators,
because that is what has been publicly reported.
With TikTok, we talked about this earlier, again in the
opening statement, nearly half of American kids use TikTok, as
you know. That is your audience. There are a lot of risks there
to privacy and national security, in my view. Ms. Pappas, I
understand that TikTok is subject to the laws of the United
States but it is also subject to the laws of other countries in
which it operates--United Kingdom, Germany.
But with regard to China, is it true, yes or no, does
TikTok have an office and employees in Beijing?
Ms. Pappas. I think this is another one for clarification--
--
Senator Portman. Just yes or no.
Ms. Pappas [continuing]. Of which TikTok does not operate
in China. You are right in saying that TikTok is subject to the
laws in the United States, as we are incorporated in the United
States and California.
Senator Portman. Do you have employees in Beijing?
Ms. Pappas. Yes, we do, as do many global tech companies,
including those----
Senator Portman. I was asking you, do you have an office in
Beijing?
Ms. Pappas. Yes.
Senator Portman. OK. Is your parent company ByteDance
headquartered in China?
Ms. Pappas. No, they are not.
Senator Portman. ByteDance is not headquartered in China?
Ms. Pappas. No. ByteDance is founded in China but we do not
have an official headquarters. It is a global company.
Senator Portman. Where is the headquarters of ByteDance?
Ms. Pappas. We are a distributed company. We have offices
around the world. Our leadership team is largely in Singapore,
but we do have an official headquarters.
Senator Portman. You have to be headquartered somewhere,
and I think it is in the Cayman Islands. Is that correct?
Ms. Pappas. The parent company was incorporated in the
Cayman Islands. That is correct.
Senator Portman. OK. You are headquartered somewhere, and
it is the Cayman Islands, but you have a presence in China, and
of course, you comply with Chinese law with regard to your
people presence in China. Correct?
Ms. Pappas. That is not correct. Again, TikTok does not
operate in China. The app is not available. As it relates to
our compliance with law, given we are incorporated in the
United States we comply with local law.
Senator Portman. Yes. Do you believe that the Chinese
Communist Party has the right to access data collected by your
company because you have a presence in China?
Ms. Pappas. Sorry, again, Senator Portman. TikTok, the app,
is not available in China.
Senator Portman. No. You said you have an office in Beijing
and you have employees in Beijing. That is a presence.
Ms. Pappas. Yes, so as we have said on the record, we do
have employees based in China. We also have very strict access
controls around the type of data that they can access and where
that data is stored, which is here in the United States. We
have also said under no circumstances would we give that data
to China.
Senator Portman. Yes. I am glad that you say that. It does
not seem to square with what we know about the Chinese national
security law, but I appreciate that approach. U.S. military
banned their own servicemembers from using TikTok for this
reason, as you know, and last month the House of
Representatives warned lawmakers of the risk of using TikTok.
These are Members of Congress were told not to use it. Our
military was told not to use it out of concern for the user's
privacy and national security.
Do you think those decisions were wrong?
Ms. Pappas. I would not opine on the needs for an
entertainment platform on Federal devices, but I would say that
TikTok is an entertainment platform first and foremost, and
this is part of the joy that we bring to millions of people
around the world. We are very much committed to the security of
our U.S. users and citizens, which is why we are investing so
heavily in this area.
Senator Portman. According to a leaked audio obtained by
Buzzfeed news, which I am sure you saw, there are TikTok and
ByteDance employees in China who can gain access to U.S. user
data, so this Committee will now be looking into the assurance
of what you said, that TikTok would not give U.S. data to
China. Do you have any response to the Buzzfeed news story?
Ms. Pappas. Yes. Those allegations were not found. There
was talk of a master account which does not exist at our
company, period.
Senator Portman. Yes. Will TikTok commit to cutting off all
data and metaflows to China, Chinese-based TikTok employees,
ByteDance employees, or any other party located in China that
might have the capability to access information on U.S. users?
Ms. Pappas. Again, we take this incredibly seriously in
terms of upholding the trust with U.S. citizens and ensuring
the safety of U.S. user data. As it relates to access and
controls, we are going to be going above and beyond in leading
initiative efforts with our partner, Oracle, and also to the
satisfaction of the U.S. Government through our work with
Committee on Foreign Investment in the United States (CFIUS),
which we do hope to share more information on.
Senator Portman. Can you make the commitment, though, that
I just asked you to make, that you will commit to cutting off
all data and metadata flows to China, Chinese-based TikTok
employees, ByteDance employees, or any other party located in
China?
Ms. Pappas. What I can commit to is that our final
agreement with the U.S. Government will satisfy all national
security concerns, yes.
Senator Portman. But you will not make a commitment to
agree to what I have now twice asked you about?
Ms. Pappas. Sorry. Given the confidentiality of CFIUS I am
not able to talk specifically about that agreement.
Senator Portman. Forget CFIUS. I am not talking about
CFIUS.
Ms. Pappas. I am happy to share more----
Senator Portman. I am asking whether you would make the
commitment today. Will you make that commitment?
Ms. Pappas. I am committing to what I have stated, which is
we are working with the United States government on a resolve
through the CFIUS process in which we will continue to minimize
that data, as well as working with Oracle to protect that data
in the United States.
Senator Portman. This is part of the United States
government too. This is our oversight function, and----
Ms. Pappas. I appreciate that.
Senator Portman [continuing]. I am concerned that you are
not able to answer the question except to say that you will not
make the commitment to cutting off this data to China. We think
that all data collected relating to Americans and then accessed
in China is a problem. We think it should be safe from
exploitation by the Chinese Communist Party. If the data is
accessible in China, as you have testified, then it could be
exploited. That concerns us.
I have gone over my time. I apologize, Mr. Chairman, but I
thought it was important to get the answers.
Chairman Peters. Thank you, Ranking Member Portman.
Senator Carper, you are recognized for your questions.
OPENING STATEMENT OF SENATOR CARPER
Senator Carper. Thanks very much. A warm welcome to our
witnesses. Mr. Mohan, thank you very much for spending time
with my staff and me earlier this week. We appreciated that.
I am a former Governor of Delaware, and the last Vietnam
veteran serving in the U.S. Senate. Not everybody knows this
but the first State to ratify the Constitution was Delaware.
Before anybody else did, we did, and for one whole week
Delaware was the entire United States of America. Those times
were a little less complex than they are today.
But I thought throughout the many years I have now lived in
Delaware a lot about our democracy, the formation of our
country and the formation of our government and how we have
rolled with the punches over many years. I never imagined how
fragile our democracy could really be.
Right after the Founding Fathers--they used to work on the
Constitution up in Philadelphia--Ben Franklin was leaving
Independence Hall, as I recall, and he was asked by someone, a
passer-by, who said, ``What have you wrought?'' he responded,
``A republic, if we can keep it.''
Churchill had his sense. He described democracy as ``the
worst form of government devised by word of man except for all
the rest.'' It is certainly a hard way to govern, and we live
it and feel it every day, and see it every day.
Jefferson, who wrote the Declaration of Independence,
largely, as you know, also wrote these words, and you already
said this, ``If people know the truth they will not make a
mistake.''
I think one of our challenges today, as we try hard to
reserve our democracy, is people do not really know what the
truth is, and they are not sure. I am not sure how to put that
genie back in the bottle, but we need to certainly try.
I have a question if I could. I would like to start off
with Mr. Cox. Mr. Cox, again we thank you for joining us today.
In your testimony you mentioned that Meta has identified more
than, I think, 1,000 militarized social movements and 270 white
supremacist organizations on your platform, and removed some
2.3 million pieces of content from Facebook that are tied to
organized hate. These statistics, which are deeply disturbing,
are only from the second quarter of this, 2022.
While I am glad that they all are tracking and removing
harmful content, these statistics are indicative of a troubling
trend of bad actors using social media to organize and mobilize
their followers.
To that end, Mr. Cox, what has Meta done to address the
larger threat of these various groups or organizations and the
content that they share? What more can and should be done?
Mr. Cox. Thank you, Senator. As I mentioned, we have
community standards which outline that there is no place for
terrorism, for militarized social movements, for violence-
inducing conspiracy networks across our family of apps. We have
350 experts, folks who work with law enforcement to identify
terrorist organizations, to identify for violence-inducing
conspiracy networks (VICNs), to make sure that we have up-to-
date information that we share with law enforcement in order to
understand, on a real-time basis, which of these networks to
prioritize and pay attention to.
We publish our results quarterly. We publish a transparency
report that outlines the prevalence of various categories of
bad content. In case it is useful, around 2 in 10,000 pieces of
content, 0.02 percent, is the prevalence of hate speech on the
platform as of the most recent report from this last quarter.
That is down from 50 percent 18 months ago. Each quarter we
have released the report over the last several years we have
been able to improve on prevalence, which is a sign that our AI
systems, our processes, our human systems, et cetera, are
improving. We believe that is the most important thing, in
addition to having a transparent report on how we are actually
doing on these numbers, so that outside experts, so that law
enforcement, et cetera, can evaluate along with us.
Senator Carper. Thank you for that. I like to say, if it is
not perfect, make it better, so keep working on it.
A question, if I could, for Mr. Sullivan, on misinformation
on Twitter. You stated, I believe, that Twitter makes it clear
in its guidelines that the promotion of disinformation is
against your platform's policies. We have seen numerous
examples, instances if you will, of users sharing this
disinformation during the coronavirus pandemic, previous U.S.
elections, and the January 6th insurrection right here at this
Capitol, just to name a few.
Could you take a moment please and explain for us what
policy changes Twitter has made in light of the rapid spread of
false information and how it has been effective?
Mr. Sullivan. Yes. Thank you for the question, Senator, and
I appreciate the historical context. It adds gravity to what we
are talking about today.
Senator Carper. Harry Truman once said, ``The only thing
new in the world is the history we forgot or never learned.''
It is pretty good. It is timely.
Mr. Sullivan. These are serious matters. We have policies
against COVID misinformation that have been evolving as we
learn from what is happening in the world, and the same for the
spread of misinformation. After our election work has been
ongoing, but most recently we have been beefing up all of our
policies against the spread of misinformation.
Then we have tried to be more proactive as well. For
example, with what we call interstitials and prompts that give
positive, valid, truthful information about things like voting,
where you can vote, when election ballots will be counted, and
things like that, so that people not only debunk false
information but can receive vetted information in a way that
feels more authoritative, so they know what is real and what is
not real.
There is always more to do, but our policies are always
evolving, and our software is always evolving to catch these
things earlier so that people do not have to report them. We
want to catch them before a person needs to even see it. Also
adding these prompts and other user interface elements to
prevent the spread of harmful information. For example, not
being able to retweet something or tweet something that is up
against that line.
We are continuing to evolve the product every day. We
acknowledge these are critical, societal issues, as you have
said. Thank you.
Senator Carper. If it is not perfect, make it better.
My time has expired. I just want to say, Mr. Chairman and
to our Ranking Member, this is a valuable and invaluable
hearing, a timely hearing, and I just applaud you holding it
and express our thanks to our witnesses for being here. Thank
you.
Chairman Peters. Thank you, Senator Carper.
Senator Johnson, you are recognized for your questions.
OPENING STATEMENT OF SENATOR JOHNSON
Senator Johnson. Thank you, Mr. Chairman. Again, I think
based on the morning panel and this panel as well we are
talking about a problem that is my true definition of a
problem, one that does not have easy solutions.
Again, what is harmful content is all in the eye of the
beholder. We all abhor violence. I certainly condemn white
supremacists. I condemn immediately, forcefully, repeatedly,
the violence on January 6th.
But I am concerned about the bias of your platforms. The
earlier panel would not give me any indication whatsoever, the
percentage of liberals versus conservative in your
organizations. I do not expect you will be any more forthright
in that.
But let me ask you this question. I know the Chairman likes
to talk about the white supremacists and January 6th. Democrats
love talking about that. What about the 570-plus riots that
occurred in the summer of 2020, 2,000 law enforcement officers
injured, $1 and $2 billion worth of property damage, a couple
of people killed in Kenosha, Wisconsin, dozens of buildings
burned, a couple dozen people also lost their lives during
those riots.
Mr. Cox, have you de-platformed, have you throttled back or
censored anybody that was involved in the organization of the
summer riots?
Mr. Cox. Senator, these were----
Senator Johnson. Give me a pretty quick answer. I have a
lot of territory to cover.
Mr. Cox. Senator, domestic terrorism and extremism and
calls for violence are against our community----
Senator Johnson. Did you censor anybody that organized the
summer riots, 570 of them?
Mr. Cox. Senator, I can look at the specifics----
Senator Johnson. Mr. Mohan, did you YouTube throttle back
or censor any of the rioters in the summer of 2020?
Mr. Mohan. Senator, our policies apply equally regardless
of who the----
Senator Johnson. Did you throttle back? Did you censor some
of the organizations that were responsible for the summer
riots?
Mr. Mohan. We would have applied our policies equally,
regardless of where the riots were, if it was a violation of--
--
Senator Johnson. Can you provide me with the names of
people you throttled back that were responsible for the summer
riots?
Mr. Mohan. Senator, I am happy to follow up on that.
Senator Johnson. Good. I appreciate that.
Mr. Sullivan, what about Twitter?
Mr. Sullivan. Twitter would have removed any incitement to
violence that was on our platform, regardless of----
Senator Johnson. Did you throttle back on organizers of the
summer riots?
Mr. Sullivan. For any case where we see incitement to
violence we would remove that content and take action.
Senator Johnson. Your CEO, Mr. Dorsey, was before the
Commerce Committee, I think it was in October 2020, and both
Senator Cruz and I asked him whether your platform, Twitter,
could impact our elections, and he denied it. Our three
witnesses earlier this morning completely disagreed with Mr.
Dorsey. They said absolutely Twitter and these platforms can.
Do you believe that Twitter can influence, in fact, our
elections?
Mr. Sullivan. I think Twitter plays an important role in
the public conversation.
Senator Johnson. It is really kind of a yes-or-no answer.
Can they impact our elections?
Mr. Sullivan. We are taking the----
Senator Johnson. Really, yes or no. I have a lot to cover.
Mr. Sullivan. As I described to the previous question, we
have put in place many actions and mitigations relating to
elections.
Senator Johnson. Can you impact the elections?
Mr. Sullivan. People try to use our platform to get
messages out regarding elections, and we are doing our best
to----
Senator Johnson. This morning I simply talked about the
fact that you censored the New York Post article about Hunter
Biden. We have polls that said that had the American public
known that we would not be in the ditch that we are right now.
Mr. Mohan, do you believe YouTube can impact the elections?
Mr. Mohan. Senator, there is an open public debate that
happens on YouTube every day, whether it is----
Senator Johnson. It is really a pretty simple yes-or-no
answer.
Mr. Mohan. Senator, YouTube is an open platform where there
is debate----
Senator Johnson. Mr. Cox, do you believe that Facebook can
impact the elections?
Mr. Cox. Senator, I think the public discussion that
happens on our platform are a part of the public discourse.
Senator Johnson. You do not think by censoring material
that specifically your management can impact the elections? Not
what is occurring on your platform, I mean, but your
management, your decisions, your censoring of information. Can
you impact the elections?
Mr. Cox. Senator, respectfully, with respect to the New
York Post story you reference we did not----
Senator Johnson. I know you throttled it back. But, again,
I am not asking about that specifically. Do you have the power?
Our earlier witness, Mr. Roetter, said, ``A small group of
people run these companies and have substantial power over
shaping the reality for billions of people.'' I mean, can you
just be honest?
By the way, our earlier panel basically, to paraphrase,
said do not believe a word you guys are going to tell us. But
can you at least be honest with the American public and say,
yes, you had that power. You can impact the elections. Can you
be honest with them or are you going to sit there and say,
``People talk about things on our platforms.''
Mr. Cox. Senator, we do think transparency about our
decisionmaking around our content, around not just terrorism
but around misinformation, we take all of these areas of
content seriously and we publish our work. We do think the
public deserves to know what our policies are and how we
enforce them.
Senator Johnson. We are going to have another round, it
sounds like. I have a lot more ground to cover. But let me just
start, and I think Senator Hawley did a great job of talking
about how Federal health agencies were in direct communication
with, in particular, I think it was Facebook, possibly Twitter.
We have also heard that, of course, Mr. Zuckerberg, or
Facebook, was contacted by the FBI as it relates to Russian
disinformation, and we covered that in the morning.
When it comes to how miserably we failed handling COVID, I
think one of the problems was the lack of robust information
using this marvelous device we call the internet. I mean,
doctors could have been testing out different theories of the
case and sharing that experience, but they were shut down. They
were censored.
I want to ask each one of you. I am 67 years old. As long
as I have been alive I have always been told if you have a
serious medical condition you really ought to seek a second
opinion because nobody has perfect information. I wish that
there would have been some modesty exhibited by our Federal
health authorities. Quite honestly, your platforms, in
acknowledging the fact that we do not have perfect information,
maybe we ought to let some information flourish. I mean people
were censored, eminently qualified doctors who has the courage
and compassion to treat COVID patients with cheap, generic,
widely available drugs.
I just want to ask the question, Mr. Cox, do you believe
that people ought to get a second opinion when it comes to
complex medical conditions?
Mr. Cox. Senator, we absolutely believe that building a
product where people have the ability to express their point of
view is critical to what we do. It is critical to what people
expect from the product and what people expect.
Senator Johnson [continuing]. Can you just answer the
question?
Mr. Mohan, do you believe you ought to get a second opinion
when it comes to complex medical conditions?
Mr. Mohan. Senator, as we all recall, when the pandemic
started it was an unprecedented event in history where science
was being created----
Senator Johnson. Maybe you ought to seek a second opinion.
I mean, do you think it is a good idea to get a second opinion,
or do you only go to one authority and put all your faith in
one authority? No other opinion is going to be valid. Is that
your belief?
Mr. Mohan. Senator, we worked with a wide variety of health
authorities in this country and all over the world.
Senator Johnson. OK. Mr. Sullivan, do you believe you ought
to go get a second opinion? If you get diagnosed with cancer
today are you going to rely on just one authority?
Mr. Sullivan. As a patient that sounds like common sense.
Our COVID information----
Senator Johnson. Are we not 330 million patients here?
But we were not allowed a second opinion, were we?
We were not allowed by your platforms for that second
opinion, and I think hundreds of thousands of people lost their
lives because you did not allow a second opinion to be
published on your platforms.
Mr. Sullivan. Our COVID misinformation policy only----
Senator Johnson. It was highly flawed, and I will point
that out in the second round of questions. Thank you.
Mr. Sullivan [continuing]. It only looked at information
that was demonstrably and widely believed to be true.
Senator Johnson. I need a second round of questions and I
can point that out, that it was not demonstrably false.
Chairman Peters. There will be a second round, Senator
Johnson.
Senator Sinema, you are recognized for your questions.
OPENING STATEMENT OF SENATOR SINEMA
Senator Sinema. Thank you, Mr. Chairman, and thank you to
our witnesses for joining us today.
Every day cartels post on social media platforms and
recruit teenagers in Arizona to act as drivers for illegal
operations. Lured by the promise of easy cash, these teens,
some as young as 14 years old, take their parents' cars to the
border and participate in smuggling and trafficking. Innocent
bystander and migrants have even died while these teens,
recruited by cartels on social media, flee law enforcement at
high speeds.
The Department of Homeland Security (DHS) must do more to
crack down on dangerous cartels' use of social media, secure
the border, and keep Arizona families safe.
My first question is for Ms. Pappas. According to my
office's conversations with Border Patrol leadership in
Arizona, TikTok is the platform that cartels use most
frequently to recruit Arizona teens. What steps does TikTok
currently take to ensure its algorithms do not promote cartel-
sponsored content, and can you tell me why have those efforts
not been more effective?
Ms. Pappas. Certainly, Senator. It starts for us with our
policies. Obviously that type of content, any illicit activity,
organized crime, including drug cartels, is strictly prohibited
from our platform.
In that regard we work with our trust and safety moderation
teams to ensure that we are detecting that content through our
technologies and also through human moderation, to remove that
content when found.
Furthermore, as a platform, we do not have the same product
features that other platforms do in terms of being able to have
that type of organized behavior. For example, we do not allow
links through direct messages or images to be sent through
direct messages. We also do not have group chat available.
Those types of behavior to help organized crime are limited
in terms of TikTok's platform. Any of the content that gets
posted on TikTok has to go through our content moderation.
Obviously, our work is never done there, but we are constantly
working to identify that content at scale and remove it when
found, and all of those numbers are also available through our
transparency reports.
Senator Sinema. A follow-up question then. If you are doing
that content moderation and reviewing each of those posts, how
is it that there are so many efforts that are successful on
TikTok to recruit young teens to assist cartels?
Ms. Pappas. We are striving to get that number to zero.
Obviously, this is a challenging area for everybody in the
industry in terms of being able to moderate our platforms, but
this is something that we heavily invest in from a technology
perspective as well as a people perspective. I am happy to look
into any of those cases. But I do know that when reports have
been sent, that content is immediately taken down.
Senator Sinema. Thank you. Mr. Cox, my next question is for
you. What is Meta doing, both on Facebook and on Instagram, to
prevent cartels from using your platforms to recruit teens
along our Southwestern Border?
Mr. Cox. Thank you, Senator. This is an important issue. It
is sad. I really appreciate your leadership on this issue. We
prohibit human trafficking. We prohibit these cartels. We work
with law enforcement to identify the names of the cartels, and
then we fan our systems to help us find instances of them
across our platforms and take them down right away.
Senator Sinema. My next question is for Ms. Pappas, Mr.
Cox, Mr. Mohan, and Mr. Sullivan, the whole panel. As Chair of
the Border Subcommittee in this Committee I believe it is
critical that each of your platforms work with the Department
of Homeland Security to identify cartel content and prevent
Arizona teens from being targeted for recruitment. I would ask
you to answer yes or no. When you discover that cartels are
using your platforms to recruit are you willing to commit to
sharing that information with the Department of Homeland
Security as quickly as possible.
Mr. Sullivan.
Mr. Sullivan. Yes, with the appropriate privacy and
oversight I believe we could do that.
Ms. Pappas. Yes. Similarly, following legal process and
privacy policy.
Mr. Mohan. Yes, Senator. We would cooperate as long as
there is a due legal process with the DHS and other law
enforcement as well.
Mr. Cox. Similarly, we would commit to that provided
privacy and legal concerns were addressed.
Senator Sinema. Thank you. Back to you, Ms. Pappas. Today's
hearing is about product development, and in the case of TikTok
there is no product more important than the, ``for you''
algorithm that offers content recommendation to users. There is
a real risk that TikTok could alter its algorithm to promote or
censor content on Beijing's behalf, whether that means
silencing voices that are critical of China or promoting
conspiracies or extremist content.
Has TikTok ever altered its algorithm or promoted or
downranked content based on the actual or perceived wishes of
the Chinese government?
Ms. Pappas. No.
Senator Sinema. In your privacy policy it says that TikTok,
``may collect biometric identifiers such as face prints and
voice prints.'' Has the biometric data of an American ever been
accessed by or provided to any person located in China, and if
not, is biometric data able to be accessed by anyone in China?
Ms. Pappas. Let me clarify because I think biometrics is
one that is a topic that is hard to define and everybody has
their own definition of what biometrics means. I will be clear
in how TikTok sees this.
We do not use any sort of facial, voice, or body
recognition that would identify an individual. There is no way
that we would be able to identify. The way that we use facial
recognition, for example, would be if we are putting an effect
on the creator's video. You are uploading a video and you
wanted to put sunglasses or dog ears on your video, that is
when we do facial recognition. All of that information is
stored only in your device, and as soon as it is applied to,
like that filter is applied and posted, that data is deleted,
so we do not have that data.
Senator Sinema. You are assured that there is no
opportunity that during the time between the use of the imprint
of the face print or voice print, and the deletion, that there
is no ability for anyone other than that device to access or
capture that information?
Ms. Pappas. That is my understanding, yes. I know it is a
technical area, so to the best of my knowledge the data is
stored on the devices and deleted immediately once you post
your video.
Senator Sinema. I would like follow-up. Neither of us are
experts on this technological issue, but I would like to get
some follow-up from those who are.
Ms. Pappas. Happy to, yes.
Senator Sinema. Thank you. Mr. Chair, I see that my time
has expired. I have a few more questions that I will submit for
the record. Thank you.
Chairman Peters. Thank you, Senator Sinema.
Senator Padilla, you are recognized for your questions.
OPENING STATEMENT OF SENATOR PADILLA
Senator Padilla. Thank you, Mr. Chair, and I want to thank
you for holding this important hearing today. The companies
testifying today offers users an unprecedented ability to
access, consumer, and distribute information. Mr. Chair, you
are right to focus our attention on how corporate product
design and investment choices influence the content that is
produced and distributed.
My first question for a couple of you is relative to
content moderation, and we have been talking about that
throughout the hearing here. Last year, Frances Haugen
disclosed that at Facebook 87 percent of all spending combating
misinformation on Facebook was spent on English language
content, despite the fact that only nine percent of Facebook's
users are English speakers. She also disclosed that trust and
safety investments for users in countries other than the United
States were abysmal.
An audit of Twitter's disinformation and misinformation
work, disclosed by Pieter Zatko, who testified just yesterday
in Senate Judiciary Committee, found that Twitter's integrity
lacked language expertise in the countries it was serving, even
though 80 percent of Twitter users are outside the United
States.
Your companies make commitments to all of your users who
are not just linguistically diverse but culturally diverse as
well.
A question first for Mr. Cox. In your testimony you state
that you have over 40,000 people working on trust and safety
issues. How many of those people focus on non-English language
content and how many of them focus on non-U.S. users?
Mr. Cox. Sure, Senator. I am happy to take your question.
Our safety and security teams are deployed to help our users
all around the world. Specifically on the question of
misinformation, which you mentioned, we have 80 fact-checkers
operating in 60 countries around the world. Those fact-checkers
are certified by independent fact-checking organizations.
In the United States we have 11 fact-checkers, six of whom
support Spanish language content. We also have partnerships
with Univision and Telemundo to connect people with Spanish
language authentic information around elections. We also offer
an election voting center in Spanish language to Americans to
help folks get authoritative information about where to vote,
that is tailored to their specific ZIP code.
Senator Padilla. I appreciate the information you are
sharing. Some of it has been in your testimony. I welcome more.
More is better. Do you have any idea of the breakdown of the
40,000 people I referenced? First of all, is that number
roughly accurate? If it is significantly higher, let me know.
If it is significantly lower, let us know. What I am looking at
is the ratio of English versus non-English. Do you have that
data?
Mr. Cox. Senator, I would be happy to follow up on the
specifics of how those 40,000 folks are broken down.
Senator Padilla. Thank you. I would greatly appreciate
that.
Mr. Sullivan, how many members of your trust and safety
team have non-English language expertise and focus on issues
outside the United States?
Mr. Sullivan. Yes, thank you for the question, Senator. As
a global company this is important to us. We have about 2,200
people working on content moderation globally. I do not have
the exact breakdown but we can take that back to our team.
Senator Padilla. Please. Ms. Pappas, how large is your
trust and safety team and how much does TikTok invest in your
non-English users, and I guess non-Western users?
Ms. Pappas. I do not have those numbers at hand but I am
happy to get back to you on those as well.
Senator Padilla. OK. Mr. Mohan.
Mr. Mohan. Senator, we have over 20,000 people that work on
content moderation all over the world. We are a global
platform, as you know, supporting a couple billion users all
over the world, and we endeavor to enforce our policies as well
as make sure that our recommendation algorithms work equally
well for all speakers, all over the world. We support dozens of
languages on our platform in all the countries that we operate.
To give you a couple of more concrete examples, here in the
United States our support across all those Four R's I described
in my initial testimony are not just about English but other
languages as well. For example, in Spanish our policies are
enforced. We serve up information panels not just in English
but in Spanish. Those relate to optics like elections, how to
vote, where to vote, et cetera, COVID-related information,
because families are looking for that content not just in
English in this country, but we recognize in a number of
different languages, including Spanish.
Senator Padilla. Thank you. I would appreciate more detail
and data from all of you.
Speaking of data, as some of you may or may not know, my
background is in engineering, so I am a big believer in data-
informed and data-driven policymaking. In reviewing your
testimony--and I appreciate some of the data that you did
provide, especially around dangerous content found in your
platforms, whether it is incomplete or desire for additional
data, let me just jump into a couple more questions.
Mr. Cox, in your testimony you say that Meta found and
removed 95 percent of hate speech content before it was ever
reported. Of the remaining five percent, how many users were
recommended the content in their news feed? Do you have data
along those lines?
Mr. Cox. Senator, I can offer data on prevalence, which
would be the amount of content that appears across the averages
of the content on our platform. For hate speech the prevalence
in our last report is 0.02 percent, or two out of every 10,000
pieces of content.
Senator Padilla. But you see where I am going, right?
Ninety-five percent is a good number. The five percent that you
did not catch before it was reported, if those recommended, 1,
2, 3, 5 times, that is one thing. If it is recommended tens of
thousands of times or more that is a different dynamic. That is
what we are trying to get at. If you do not have the data at
your fingertips, a follow-up would be welcome.
Mr. Cox. Senator, we would be happy to follow up on that.
Senator Padilla. Great. Ms. Pappas, in your testimony you
say that 88.4 percent of removals under TikTok's violent
extremism policy occurred within 24 hours of being posted.
Again, a good number but it is not 100. For the other 11.6
percent, do we have a gauge of how long it took to find and
resolve those items?
Ms. Pappas. No. I would have to get back to you on that,
but similarly we look at the prevalence of content
authoritarian would be violative, and for violent extremism it
is 0.01 percent.
Senator Padilla. OK. Thank you.
Mr. Chairman, my time is up. Similar to Senator Sinema I
will have some additional questions I will submit for the
record.
Chairman Peters. Very good. Thank you, Senator Padilla.
Senator Hawley, you are recognized for your questions.
OPENING STATEMENT OF SENATOR HAWLEY
Senator Hawley. Thank you very much, Mr. Chairman. Thanks
to all the witnesses for being here. Ms. Pappas, let me start
with you.
I have to say it is great to see you here today. I have
repeatedly invited your company to testify before Congress. I
invited them to testify to the Judiciary Subcommittee on Crime
and Terrorism in November 2019. I invited them to testify again
in September of the following year. Both times we were stiffed.
TikTok told me that they would set up a meeting with the CEO.
They did not want to testify in public but they set up a
meeting with the CEO after November 2019. They then canceled
that meeting.
It nice to see TikTok be willing to answer questions in
public. It is a pleasant change. Let us dig into a few things,
if we could, specifically about TikTok's links to the Chinese
Communist Party.
In response to a letter from some of my colleagues, TikTok
claimed earlier this year that the company has never shared
data with the Chinese government. Is that correct?
Ms. Pappas. That is correct, yes.
Senator Hawley. And has never shared data with the Chinese
Communist Party. Is that correct?
Ms. Pappas. We will never share data, period.
Senator Hawley. My question was in the past tense. Has
TikTok ever shared data with the Chinese Communist Party?
Ms. Pappas. We have never shared data with the Chinese
government. Correct.
Senator Hawley. With the Chinese Communist Party.
Ms. Pappas. Yes, correct.
Senator Hawley. Have you ever shared it with members, to
members of the Chinese Communist Party?
Ms. Pappas. We have said many times, Senator, that we do
have Chinese engineers based in Chinese. I do not think there
is any platform up here that would be able to speak to what you
are talking about as it relates to the political affiliation of
an individual. But I am happy to assure you that we are ensuing
the access controls around our data as well as the storage of
that data in the United States.
Senator Hawley. I think you are telling me that there are
TikTok employees or ByteDance employees who are members of the
Chinese Communist Party. Is that what you are saying?
Ms. Pappas. No. I am saying I would not be able to verify
that.
Senator Hawley. Let me ask you affirmatively. Are there
TikTok employees or ByteDance employees who are members of the
Chinese Communist Party?
Ms. Pappas. Senator, I am saying that nobody that is
sitting on this panel could tell you a political affiliation--
--
Senator Hawley. I am not interested in anybody's opinion. I
am asking you a factual question. Are there members of the
Chinese Communist Party employed by TikTok and ByteDance? Yes
or no.
Ms. Pappas. I would not be able to tell you the political
affiliation----
Senator Hawley. You do not know?
Ms. Pappas [continuing]. Of any individual. What I can tell
you is how much we are investing in the----
Senator Hawley. No. Membership in the Chinese Communist
Party is not exactly like membership in the Democratic Party. I
am looking for an answer. You are telling me you do not know?
TikTok does not know.
Ms. Pappas. Here is what I can tell you. I can tell you
that in our United States and Singapore leadership, there are
no CCP members.
Senator Hawley. You do know that. But you are telling me
that you do not know if there are any members who are employed
by TikTok or ByteDance, members of the Chinese Communist Party?
Ms. Pappas. Senator, I am happy to share that we are
putting access controls----
Senator Hawley. That is not my question.
Ms. Pappas [continuing]. As well as----
Senator Hawley. That is not my question. My question is are
there any TikTok employees or ByteDance employees, members of
the Chinese Communist Party? Yes or no.
Ms. Pappas. Senator, I am saying nobody could sit up here
and give you that.
Senator Hawley. You are saying you do not know? But you do
know your leadership is not but you do not know about your
employees. Is that your testimony?
Ms. Pappas. I know that everyone who makes a strategic
decision at this platform----
Senator Hawley. Yes.
Ms. Pappas [continuing]. Is not a member of the CCP.
Senator Hawley. A strategic decision. OK. It is
interesting. It is interesting to me that you are quite
confident that anyone who could make a strategic decision--how
many people is that?
Ms. Pappas. It is our leadership team.
Senator Hawley. The number?
Ms. Pappas. Again, the leadership team is based in the
United States and Singapore. Our CEO is based in Singapore. He
is not Chinese. I am happy to go into the efforts that we----
Senator Hawley. Would it surprise you to learn that Forbes
magazine recently reported that at least 300 current TikTok or
ByteDance employees were members of Chinese State media and
affiliated with the Chinese Communist Party?
Ms. Pappas. Again, we do not look at the political
affiliations or cannot speak to individuals, but what I can
tell you is that we are protecting the data in the United
States.
Senator Hawley. But apparently, though, you do look at
political affiliation because you are quite willing to sit here
and tell me that no one who has strategic input or makes
strategic decisions is a member of the Chinese Communist Party.
You do know very well, as a matter of fact. You just do not
want to answer my other question.
Ms. Pappas. We have thousands of people that work at the
company so I am not going to vouch on the political affiliation
of any particular individual. What I can vouch for----
Senator Hawley. Have you seen the videos of Chinese
Communist Party members conducting training for TikTok and
ByteDance employees?
Ms. Pappas. No.
Senator Hawley. That is fake?
Ms. Pappas. I do not know what you are referring to. But
what I can tell you----
Senator Hawley. Has that happened?
Ms. Pappas [continuing]. Is any decision----
Senator Hawley. Has that happened?
Ms. Pappas. Sir, I just said that I would not be able to
tell you. I have not seen it. I am not sure what you are
referring to, but I am happy to follow up. But what I can tell
you----
Senator Hawley. Wait. I am sorry. Let us go back. Let us
see if we can cut through the mumbo-jumbo. I am asking you if
the Chinese Communist Party has conducted training sessions
ever for employees of ByteDance or TikTok. Yes or no.
Ms. Pappas. Not for TikTok. TikTok, the app, does not
operate in China.
Senator Hawley. You have employees in China and ByteDance
has employees in China. Listen, we have been through this song
and dance.
Ms. Pappas. We have.
Senator Hawley. Let us just skip that. I have heard it all
before.
Ms. Pappas. Senator, I appreciate----
Senator Hawley. Answer my question. Yes or no. Have they
conducted training for ByteDance employees or TikTok employees.
Ms. Pappas. I can speak on behalf of TikTok, and the answer
is no.
Senator Hawley. No. That is interesting. Do any TikTok
employees based in China have access to U.S. user data?
Ms. Pappas. As we have publicly said, yes, we have
engineers in China, and we are working on the access controls--
--
Senator Hawley. None of them are members of the Chinese
Community Party?
Ms. Pappas [continuing]. We are working on the access
controls to minimize that data access----
Senator Hawley. I have heard that, and frankly I do not
believe it.
Ms. Pappas [continuing]. Working with the United States and
through the CFIUS----
Senator Hawley. Wait. So your testimony is that you do have
TikTok employees based in China who do have access to U.S. user
data, but you are confident that none of them are members of
the Chinese Communist Party and they never accessed it? Is that
your testimony?
Ms. Pappas. Anyone who has access to U.S. user data has and
does so to perform daily duties, so if it is for the
performance of site management, bug handling. But we have
strict controls in terms of who and how our data is accessed.
Senator Hawley. None of that is accessible to any member of
the Chinese Communist Party. Is that your testimony?
Ms. Pappas. We believe we have the strictest controls out
there----
Senator Hawley. That is not my question.
Ms. Pappas [continuing]. Actually we are working with
Oracle----
Senator Hawley. My question is does anyone who has access
to user data, are they members of the Chinese Communist Party?
Ms. Pappas. I feel like I have answered your question.
Senator Hawley. You have not, and I feel like you are
avoiding it----
Ms. Pappas. No.
Senator Hawley [continuing]. At every opportunity.
Let me give you another one, since you are on the record
and under oath.
Ms. Pappas. Can I be as clear as----
Senator Hawley. I would welcome you being clear.
Ms. Pappas. Thank you.
Senator Hawley. Does any person who has access to U.S. user
data, are they members of the Chinese Communist Party? Yes or
no.
Ms. Pappas. Let me be clear again.
Senator Hawley. Yes or no.
Ms. Pappas. For our U.S. users, the data is sorted and
housed in the United States. We have access controls in place--
--
Senator Hawley. You are not answering my point. Let the
record reflect you will not answer my question. Why not?
Ms. Pappas. Any of that data, it is overseen by our U.S.
led security team.
Senator Hawley. That is not my question.
Ms. Pappas [continuing]. And monitored daily.
Senator Hawley. That is not my question.
Ms. Pappas. Furthermore----
Senator Hawley. My question is does any employee who has
access to U.S. user data, are they members of the Chinese
Communist Party? You will not answer that.
Ms. Pappas. Again, as a global technology platform there is
no other company that could make that assertion either.
Senator Hawley. That sounds like a yes to me. I think that
is news.
You are familiar, I know, with this Buzzfeed article that
says that according to leaked audio at more than 80 internal
TikTok meetings, China-based employees at ByteDance have
repeatedly accessed non-public data about U.S. TikTok users.
``Everything is seen in China,'' said a member of China's Trust
and Safety Department in a September 2021 meeting. In another
September meeting a director referred to one Beijing-based
engineer as a ``Master Admin who has access to everything.''
These reports show data was accessed far more frequently and
recently than previously reported. Your testimony is that this
is false?
Ms. Pappas. Correct.
Senator Hawley. All of this is false.
Ms. Pappas. That is correct. Everything that you just
stated, there is no such thing as a Master Account.
Senator Hawley. That is not what it says. It says that
someone is referred to as ``Master Admin.''
But you are telling me that China-based employees have
never accessed non-public data of U.S. TikTok users.
Ms. Pappas. No. I have already said on the record that we
have Chinese employees who have accessed data.
Senator Hawley. That is what this is saying. So you agree?
Ms. Pappas. If you want to clarify on each individual
statement. I am saying that there are strict access controls
around the data that is accessed in the United States. That is
overseen by our U.S. led security team. We are working with
Oracle.
Senator Hawley. That is not what this article says.
Ms. Pappas. We disagree with the categorization in that
article, wholeheartedly.
Senator Hawley. Here is the point. I know there are other
Senators who want to ask questions. I think we are going to
have a second round. The truth appears to be, besides the fact
that we cannot get a straight answer on any of these questions,
is that you have hundreds of employees with, it appears, access
to U.S. user data, that may very well be members of the Chinese
Communist Party. You have no way to assure me that they do not
have access to our citizens' data. You will not answer my
question in a straightforward way about whether a CCP has ever
gained access or not.
I think, for my own point of view, that is a huge security
problem.
Ms. Pappas. Senator, if I may. We are one of the most
highly scrutinized platforms. There have been many
cybersecurity experts who have researched our platforms,
including Citizen Lab, which is a leading academic research
unit based in the University of Toronto, who have said, and I
am happy to submit this for the record for the Committee, that,
``Our research shows that there is no overt data transmission
to the Chinese government by TikTok.''
Senator Hawley. Overt.
Ms. Pappas. ``TikTok's features and codes do not pose a
threat to national security.''
Senator Hawley. Wait a minute. Overt data transmission?
Ms. Pappas. There are also----
Senator Hawley. Ms. Pappas, this is not a hearing for you
to testify at will. You are here to answer questions.
Ms. Pappas. I am providing you with information.
Senator Hawley. No, you are not. You are talking over me,
and you are submitting information from--who knows who funds
this entity, who knows who is behind it, who knows what it
contains? I do not know.
What I do know is you will not give me straight answers to
my questions, and the reason, I think, is pretty clear, because
your company has a lot to hide. You are a walking security
nightmare. For every American who uses this app, I am
concerned.
Chairman Peters. Senator Hawley, thank you.
Senator Ossoff, you are recognized for your questions.
OPENING STATEMENT OF SENATOR OSSOFF
Senator Ossoff. Thank you, Mr. Chairman, and thank you to
our witnesses today.
Mr. Sullivan, in disclosures he has made publicly and to
the Congress and in his testimony yesterday, former Twitter
employee, Mr. Zatko, alleged that Twitter has made willful
misrepresentations to the Federal Trade Commission with respect
to its compliance with past regulatory action. Is that true?
Mr. Sullivan. I am familiar with the allegations. I would
point you to our statements that we made as a company that the
company disagrees with much of the allegations. Now it is
connected to an ongoing lawsuit, so I am not able to----
Senator Ossoff. My question to you, Mr. Sullivan, is has
Twitter willfully misrepresented facts to the Federal Trade
Commission?
Mr. Sullivan. I can tell you that Twitter disputes the
allegations, is all I can tell you about those particular
allegations.
Senator Ossoff. You cannot tell me definitively, Mr.
Sullivan, that Twitter has not willfully misrepresented facts
to the FTC.
Mr. Sullivan. I would point you to what I just said.
Senator Ossoff. Noted. You do not deny that Twitter has
willfully misrepresented facts to the FTC. Understood.
I want to ask you about the logging of access to user data
and the extent of privileged access to user data for Twitter
personnel. Does Twitter, Mr. Sullivan, have in place a system
by which you can determine definitively which Twitter employees
have accessed private user data, for example, to include
history of use of the platform, browsing history, direct
message, geolocation data, Indo-Pacific addresses?
Mr. Sullivan. Thank you for the question. I can tell you
what I have observed. I have been in my role since April of
this year. Our current leadership for infosec, the privacy and
access controls has a robust process for access to data.
For example, people have to have a business need to access
certain datasets, so we have to operate the service, some
number of people need access to certain datasets. Our goal,
that is aligned with our privacy objective, is to minimize that
access to that necessary to do your job function.
We have access controls, monitoring, logging. I receive,
for example, new employee approvals, this person needs to be
able to run this report.
Senator Ossoff. Mr. Sullivan, I appreciate the overview,
but the specific question to which I am seeking an answer is,
is there a log event any time a Twitter employee accesses the
private user data of a specific user? Can Twitter determine
every time one of your employees has accessed such private user
data? Do you have that functionality? It is really a yes-or-no
question.
Mr. Sullivan. We have monitoring and logging and access
control. It is always evolving and improving. But what I can
tell you is I have observed it in action. I cannot speak to
every single system. We have a team that can.
Senator Ossoff. Mr. Sullivan, I will look for that in the
follow-up. I want to say, respectfully, this. You are here
before the U.S. Senate. Serious allegations were made yesterday
by one of your former employees, and I am open-minded. I am
here pursuing the facts. Certainly in your responses for the
record it is going to help you to be clear, definitive, and
precise responding to yes-or-no questions like that one. Can
you commit that in your written responses we are not just going
to get talking points and generalities, we are going to get
precision and yes-or-no answers to yes-or-no questions?
Can I get a yes-or-no answer to that question?
Mr. Sullivan. Yes. I am trying to explain----
Senator Ossoff. Thank you. No. I just need a yes to that
question.
Mr. Sullivan. Yes, I understand. Thank you.
Senator Ossoff. So yes?
Mr. Sullivan. Yes.
Senator Ossoff. Great. Thank you.
Let me ask you, please, Mr. Cox. There has been substantial
public reporting controversy and concern about the Metapixel
product and the possibility that its deployment on various
hospital system websites, for example, has enabled Meta to
collect private health care data, some of it potentially that
would typically be Health Insurance Portability and
Accountability Act of 1996 (HIPAA) protected, from U.S.
persons.
Does Meta possess or collect any health care or medical
data related to its users or to U.S. persons?
Mr. Cox. Senator, not to my knowledge, but I would be happy
to follow up on that specific issue.
Senator Ossoff. OK. I would like you to follow up, and
please, Mr. Cox, submit to this Committee a comprehensive and
precise answer to that question, which I will recharacterize in
writing. We need to understand, as the U.S. Congress, whether
or not Meta is collecting, has collected, has access to, or is
storing medical or health data for U.S. persons or your users.
Will you get me a comprehensive and precise answer to that
question?
Mr. Cox. Senator, yes, we would be happy to follow up.
Senator Ossoff. OK. Thank you very much.
Ms. Pappas, I overheard some of the responses to Senator
Hawley's question. I would like you to answer a question. There
has been a significant topical focus on this throughout this
hearing. In what ways does the government of the People's
Republic of China, if at all, exercise influence over TikTok's
corporate behavior or corporate policies? I am going to ask the
Chairman's indulgence and follow-up for as much precision as I
can get, so I am going to humbly and respectfully ask you not
to give me the immediate topline talking points but to give me
a precise, particularized answer to that question.
Ms. Pappas. In no way, shape, or form, period.
Senator Ossoff. In no way, shape, or form, period, does the
government of China exercise any influence over TikTok's
corporate practices or policies.
Ms. Pappas. Correct.
Senator Ossoff. For example, if you receive a response from
the government of China to take down certain content for
reasons that they State are related to their national security,
do you comply with such requests?
Ms. Pappas. No.
Senator Ossoff. Do you comply with such requests if you
receive them from the U.S. government?
Ms. Pappas. If it follows due legal process, yes. We
actually include all government requests for takedown in our
transparency reports, in which you can see that China has not
requested.
Senator Ossoff. Thank you, Ms. Pappas. There will be some
follow-up questions for you there for the record. I appreciate
all of your testimony. Thank you for answering questions, for
those which were answered, and Mr. Chairman, I yield back.
Chairman Peters. Thank you, Senator Ossoff.
Senator Lankford, you are recognized for your questions.
OPENING STATEMENT OF SENATOR LANKFORD
Senator Lankford. Mr. Chairman, thank you. Thank you to all
of you and your testimony. You have been here a long time.
There are a lot of questions. You have gone through a lot of
different issues. I apologize I had to be able to run in and
out real quick.
Ms. Pappas, I want to be able to follow up on a couple of
things real quick. You have answered a lot on China. Obviously
it has been a big issue. You know that. It is not like you went
to TikTok and were shocked there were issues with China and the
possibility there.
There are a couple of questions that have come up recently
on this. One of them is the ability for TikTok to be able to
track keystrokes after you leave the app, to be able to be on
the app, click a link to be able to go to another site, and be
able to track keystrokes. Is that a part of the app's design
that you can do that?
Ms. Pappas. No, it is not.
Senator Lankford. That is not used, because it has been
widely reported that is part of the app currently and its
structure. Has it been part of the app and has it recently been
taken off?
Ms. Pappas. The keystrokes one, to my knowledge, was
basically an anti-spam measure, and so that was never
collecting the content of what was being typed.
Senator Lankford. Was there an ability, though, to be able
to track keystrokes on it as you are on the app, click on a
link to be able to go to another page, to be able to track?
Ms. Pappas. I do not believe so, no.
Senator Lankford. OK. We will follow up on that ``do not
believe so'' on that.
The other one is you have offices all over the world. As
you mentioned, a lot of your offices are in Singapore. The
original development of TikTok, it is my understanding it came
from ByteDance. It was a Chinese development originally and
then it spread all over the world. Correct?
Ms. Pappas. Yes, it was originally developed by the parent
company, ByteDance, but also Musical.ly, the app, the two were
combined. But currently, and for a while now, there have been
separate apps, separate codes, separate servers.
Senator Lankford. Are any of the developers that still work
on the design still based in China, in your Chinese office?
Ms. Pappas. Yes. We have said that we have engineers in
China. Correct.
Senator Lankford. That will be one of the conversations we
will have in the days ahead to be able to follow up on, what
those access points are. There is an obvious consideration here
with this Committee and with others on national security
issues. It is just well known that China has, as a part of
their law, they get access to anything with technology. For
China to have the possibility to have access to 100 million
Americans, including most of our young people, that is an issue
for us, and it is the reason we ask hard questions.
Ms. Pappas. We understand that concern and I appreciate
your question, Senator, which is why we are investing heavily
in ensuring strict access controls, and we are working with
Oracle. We recently announced that 100 percent of our user data
is now stored in Oracle's cloud infrastructure, and we have
further said that they will be vetting and validating our
content moderation and recommendation systems. We really are
committed to transparency and security on these topline issues,
and we are happy to provide further information.
Senator Lankford. Great. We will continue to be able to
follow up.
Mr. Cox, thank you, as well, for being here, as for all of
you in this conversation. I have a couple of questions here.
One is dealing with the experts, as you mentioned in your
testimony as well, that are actually helping with the fact-
checking process. We did a little bit of digging in some of
this, and obviously you have a diverse group of nonprofits and
think tanks and other folks that help some of the experts in
fact-checking. But there are also some that make us scratch our
head a little bit on it.
There was one of the groups that was dealing with
coronavirus and some of the fact-checking early on on that was
actually a group of journalists. As we went through and looked
at some of the credentials, all of which were public on all
these individuals--thanks for the transparency on that--none of
them were medical professionals on it.
Not to be pejorative on journalists, but I do not run into
a lot of conservative journalists. There are a few out there.
The consistent fear is that conservative voices are silenced,
and when I look at some of the groups that actually do the
fact-checking I do not find a lot of conservative groups that
do this.
Ms. Pappas, on the same kind of issue, as I go through for
TikTok they list as one of the fact-checking groups, or the
experts that are out there, the Southern Poverty Law Center as
one of the places they go. The Southern Poverty Law Center is
considered the Family Research Council and the Alliance
Defending Freedom, which are just pro-family groups and
religious freedom groups, as hate groups. If TikTok is
dependent on the Southern Poverty Law Center to be able to find
what is a hate group, the Family Policy Council is a hate
group, suddenly, on TikTok.
The question is, how do you develop your expert groups? How
do you make sure that they are actually balanced and that the
advice you are getting on what that looks like is actually
fair?
Mr. Cox, do you want to jump in first on the Meta side?
Mr. Cox. Yes, I would be happy to, Senator. On the issue of
misinformation, we know that people do not want misinformation
on the platform, and that is why we have developed a program to
work with independent fact-checkers that are certified by the
Independent Fact-Checking Network (IFCN).
Senator Lankford. How do you make sure it is a balanced
perspective, philosophically?
Mr. Cox. I know that the IFCN has specific policies around
looking for balance. I also know that there are folks on both
sides of the aisle who are members of that network.
Senator Lankford. I would only say, how do you make sure it
is balanced, not that organization make sure it is balanced,
because again, there is a perception--and I would tell you, I
understand their perception because I have a lot of
conservative organizations--churches, faith-based nonprofits,
all kinds of entities--that reach out to my office, at home and
here, and they will reach out to me and say, ``I just got
blocked from Facebook. We are trying to figure out why.'' They
are not terrorist organizations. They are not violent. They are
not anything else. They just got blocked, and they are trying
to figure out if conservative ideology is the reason why.
What I am trying to figure out is who fact-checks the fact-
checkers for you to be able to make sure that you are getting a
fair perspective on this? You have millions or billions of
pages IFCN needs to be able to track on this. When someone
gives you counsel, how do you take advantage of checking that
first to make sure it does not have a bias?
Mr. Cox. Senator, on the question of checks and balances
among the fact-checker network, the system that we have set up
allows fact-checkers to check each other and resolve dispute
claims in that way. We believe that that helps the system be
more fair.
Senator Lankford. But I guess I am asking, so the same
question. How do you make sure that that perspective is
balanced, that it is not all fact-checkers that think alike?
Mr. Cox. Senator, ultimately we believe that our platform
is best for people when it can be a place for all voices and
for all political points of great.
Senator Lankford. Great.
Mr. Cox. That is in our interests and that is in the
interests, we believe, of the Nation.
Senator Lankford. I 100 percent agree. I am trying to say
to you it is not, that there are entities that really believe
their voices are being blocked out, and that the individuals
that are fact-checking them have a bias against them
politically, not necessarily for violence or something else.
That is part of the challenge here, that I would challenge you
on that, and for all of us, to be able to make sure it is going
to be fair and balanced on that.
Let me move on to a couple of other issues. Mr. Chairman,
do I have an extra minute here I can go on it? Thank you for
that. I have two quick other things that I want to be able to
address. In Meta's terms of service you state, in terms of
service, 3.2.1, ``You may not use our products to do or share
anything that is unlawful, misleading, discriminatory, or
fraudulent, or assists someone else in using our products in
such a way.''
But you also have stated, as Meta, ``We prohibit content
that offers to provide or facilitate human smuggling, which
includes advertising a human smuggling service, but we do allow
people to share information about how to enter the country
illegally or request information about how to be smuggled.''
Now I am trying to align those two, where you say you
cannot use our platform for any illegal activity, promoting
illegal activity, facilitating that unless you are illegally
crossing our border. Then we are going to facilitate the use of
our platform, and in fact, has been accessibly used to
facilitate connecting with the cartels and the traffickers to
be able to facilitate illegally crossing our Southern Border.
Help me understand between those two. Which one is correct?
Mr. Cox. Senator, we have been working with law enforcement
for a while now on the very serious issue of human trafficking
across our borders. We have folks at the company who
specifically speak to law enforcement and border officials to
make sure we have an up-to-date list of which cartels that we
can use in order to fan out our systems and make sure we take
them down. We have policies against human trafficking, and we
have policies against those cartels, to make sure that we are
able to remove them as soon as they pop up.
Senator Lankford. But this is a Meta statement that you
said, ``We allow people to share information about how to enter
the country illegally or request information about how to be
smuggled.'' That is allows already, based on Meta's policy, but
you also say, ``We do not allow you to use this for illegal
activity.'' That is what I am trying to figure out, is you
either allow for illegal activity or you do not allow for legal
activity, but it looks like you are trying to do both. We do
not like smuggling, but we are facilitating people who are
illegally coming into the country.
Mr. Cox. Senator, the policy here, as I understand it, is
specifically about human trafficking and cartel networks that
are facilitating illegal trafficking of people. I would be
happy to follow up with you on this.
Senator Lankford. Let us do this. There is not a person
that crosses our Southern Border into the United States that
does not pay the cartel. As our Border Patrol will tell you,
the border is secure. It is secure on the south side.
When I was in McAllen, Texas, a couple of months ago they
said in that area the cartels, just in that area, make $153
million a week trafficking people across the border. Many of
those meet up with those people that are moving them across the
border illegally, through a Facebook platform. That is a big
issue to me, and it seems like Meta is being inconsistent in
their terms of service about illegal activity.
My last comment on this, and I really will make it my last
comment, and I really do appreciate the time on it, I have, for
years gone back to Facebook and said, ``I have all kinds of
constituents at home that say to me, `I would comment to you on
your page except when I comment I get just ruthlessly attacked
by people that politically disagree.' '' They click the angry
button, they yell at them, they say all kinds of mean things to
them when they comment on my page. So they just do not comment.
What has happened is, political Facebook pages--and that is
for everyone here, both sides of the aisle--have become places
for anger and aggression. When you disagree you go and attack
people that comment, that like someone politically, you go
attack them instead.
What I have asked Facebook for years is, allow those of us
that we know our pages are place where there is wide
disagreement to be able to have the option to say, ``You can
comment to me but you cannot attack the people that comment to
me.'' We can have dialog and interaction but you cannot have
this angry interaction with each other on the page. Give us the
option to turn that off so we see comments, we can respond back
to people and have that dialog and interaction, but you turn
down the volume.
What I have heard, year after year, from Facebook is,
``That is not really what we do. What we do is interaction.''
But everyone knows the interaction there is angry, bitter,
aggressive interaction. That is not healthy.
My request again to Facebook, which we have done in writing
and in follow-up and in conversation, is you have the ability
to turn down the volume, to be able to have fewer angry emojis
flying at people, by allowing that interaction. Please do.
Chairman Peters. Thank you, Senator Lankford, for the
Senate one minute. I appreciate it. [Laughter.]
We will do a second round, and because of the late nature,
and all four of you have been here a long time, if everyone can
hold within--I was very generous in the first round. We will
try to make sure that we do the seven minutes in this round, if
you would.
I want to get back to where we left off which was on the
actual design of these products up front, not dealing with
problems after they have already arisen and sometimes waiting
years before you fix the problems. How do we put it in the
initial design?
My question, Mr. Cox, is Wall Street Journal (WSJ) reported
last week that Meta shut down its Responsible Innovation Unit.
These two dozen employees were charged with identifying
potential harms at the conceptual stage of new product design
and change the design culture. Why was that eliminated?
Mr. Cox. Senator, thanks. Respectfully on this team,
because I saw this reporting as well, the work here was not
eliminated. The specific team named here was a small team of
about 20 people that was overlapping with our much broader
integrity, safety, civil rights efforts across the company.
This was a case of just moving that work into the teams where
it was best----
Chairman Peters. You just moved them into a different part.
They are still there. Is it safe to say that this is happening
across your platform, with our design team? Is everybody on the
design team, are they compensated based on the trust and safety
of the products that you are putting out? Is that part of the
metric?
Mr. Cox. Senator, when we look at how health of any product
we will look at trust and safety as a part of that. We will
look at security. We will look at relevance. We will look at a
holistic set of metrics, both quantitative and qualitative.
Chairman Peters. No, I know. Let me interrupt because of
time, because I am going to hold everybody to seven minutes
here. In the time, with the metric to determine compensation,
does an individual in that metric, do they get compensated
based on something related to safety and trust. When they get
their bonus at the end of the year, are safety and trust part
of that compensation?
Mr. Cox. Senator, so for bonuses we would have----
Chairman Peters. Just say yes or no.
Mr. Cox. Is it a part of how we look at the health of the
product, which is related----
Chairman Peters. It is a part. It is an actual line.
Mr. Cox. Excuse me, Senator?
Chairman Peters. It is an actual line, related to safety
and trust.
Mr. Cox. Trust and safety metrics are part of----
Chairman Peters. If a product goes out that causes a lot of
problems, they are going to be penalized for that, financially?
Mr. Cox. Senator, we would not launch a product if we
believe it was about to be unsafe.
Chairman Peters. You would not.
Mr. Cox. Once we do launch products we evaluate things like
prevalence, things like reports, and a whole host of metrics in
order to understand the health of a product from a safety
perspective.
Chairman Peters. But do people get compensated related to
safety and trust? Just yes or no. You said yes, they do. I will
go down. Mr. Mohan.
Mr. Mohan. Senator, building in trust and safety into our
products is not just an integral part of our goal, so topline
metrics, it is our No. 1 priority. But it is also built into
the product development process.
Chairman Peters. I know it is in the process. I just want
to know the compensation. Are they compensated specifically
because they are working on trust and safety? Is every employee
in your product team doing that?
Yes or no.
Mr. Mohan. If an employee builds a product that does not
factor into account trust and safety we simply would not launch
that product for our users.
Chairman Peters. We heard today that people had questions
about launching of products, and they still got launched, that
trust and safety folks--thank you for your opinion but we have
to launch this product because people are compensated--second
question, they are compensated based on growth and
profitability, like other companies do. You are not the only
company that does that, based on growth and profitability, but
that is really the main driver.
Mr. Mohan. Senator, as the Chief Product Officer of YouTube
I look after both our product development process and our trust
and safety operations, and I can tell you, unequivocally, that
we would not launch a product or grow a product that was at the
detriment of our users' trust and safety.
Chairman Peters. If you launched a product and it turned
out it was not like you thought, and it was not trustworthy or
safe, would the product designer lose their bonus or they would
be compensated?
Mr. Mohan. It would factor into their performance reviews.
Chairman Peters. OK. Ms. Pappas. Yes or no.
Ms. Pappas. Safety and trust is a core priority for us.
Chairman Peters. I understand. Is it a part of
compensation?
Ms. Pappas. Every trust and safety, like every feature,
rather, has trust and safety as a seat at the table. As we do
our product development and launch process we have actively
delayed launches based on not meeting the merits of safety. It
is a top priority for us. We invest heavily in this to ensure
the safety of our products at launch. In regard to performance,
it is one of the factors.
Chairman Peters. Mr. Sullivan? Be brief, please. Yes or no.
Mr. Sullivan. The health and safety is a topline metric for
the Consumer Products organization, so that will affect how
people's performance is graded.
Chairman Peters. You have mentioned examples of where you
may not have launched or have not launched. I assume that you
all have examples where you have not launched because some
issues were raised. I would certainly like to have that
information. Would each of you commit to giving us an example
so we have a sense of what actually is caught before it is
actually released?
Mr. Cox, not now, but would you provide an example for us
on that?
Mr. Cox. Senator, we would be happy to.
Chairman Peters. Thank you. Mr. Mohan.
Mr. Mohan. Yes, Senator, I am happy to follow up.
Chairman Peters. Ms. Pappas.
Ms. Pappas. Yes.
Chairman Peters Mr. Sullivan.
Mr. Sullivan. Yes. I also have examples.
Chairman Peters. Great. Thank you. The other thing is how
you deploy resources, and we have heard a lot of numbers here.
I think the most valuable resource is just the number of
engineers. I am going to go and ask you three questions, each
of you to answer.
We sent this to you last week. We have been trying to get
this information for a long time. We said I am going to ask you
this question today, so I am sure you are prepared for the
question because we asked it Friday.
Each of you, what is the total number of full-time
engineers you have in your company? How many of those engineers
work full-time on ensuring trust, safety, or integrity of your
platforms? Three, how many engineers work full-time on product
development?
Mr. Cox.
Mr. Cox. Senator, the total number of engineers at the
company is on the order of tens of thousands.
Chairman Peters. No. That is not what I asked. We asked
very specific questions on Friday. We have been trying to get
this information for a long time. We said we are going to ask
you this question in the hearing, and you are saying you did
not get it. You do not have it for me? OK. Mr. Mohan.
Mr. Mohan. We have thousands of engineers that work at
YouTube.
Chairman Peters. OK. You do not have a specific answer for
me either. Ms. Pappas.
Ms. Pappas. I do not have the engineer numbers but trust
and safety represents our largest labor expense for TikTok.
Chairman Peters. OK. You do not have numbers. Mr. Sullivan,
you do not have numbers as well, or do you? I hope you do.
Please, one of you do. We have been trying for months to get
these answers. This is why we get so frustrated.
Mr. Sullivan. We have about 2,200 people working on trust
and safety across Twitter.
Chairman Peters. What is the total number of full-time
engineers?
Mr. Sullivan. I am sorry. That was not an engineer number.
This is those that build and enforce the Twitter rules. We have
several thousand engineers at Twitter.
Chairman Peters. So the same thing. You do not have
specific numbers, as we asked. OK.
Would you commit to get me those numbers, Mr. Cox?
Mr. Cox. Senator, I am happy to have the teams follow up.
Chairman Peters. That is a yes. Mr. Mohan.
Mr. Mohan. Senator, I will have my teams follow up as well.
Chairman Peters. Thank you. Ms. Pappas.
Ms. Pappas. We are actively working to get you those
numbers. We will follow up as appropriate, yes.
Chairman Peters. Thank you. We are trying to work together.
This is really a complex problem. We get it. I understand the
complexity of the problems you have to deal with each and every
day. We want to work with you, but we need to be able to have
this kind of dialog to get a better sense of what is going on
as we go forward, so please do that.
Ranking Member Portman, you are recognized for your
questions.
Senator Portman. Thank you, Mr. Chairman. Not to leave
Twitter out, I wanted to ask a question regarding the sexual
material online that we talked about earlier. As I said, this
Committee has been a leader in stopping human trafficking, and
specifically sex trafficking underage kids, and we have passed
some legislation that is making a difference.
Based on a website called Bark, that advises parents on how
to get their kids safe, among the top five severe sexual
content sites was Twitter. This year it was widely reported
that Twitter considered monetizing sexual content, meaning, as
I understand it, people could actually get paid for
pornography, basically, for putting sexual content online. My
understanding is this project has now been put on ice because a
group of Twitter employees found that the platform could not
effectively separate out child exploitation content, and I
appreciate you did not go forward with this plan.
According to Verge, the Twitter employees have said that
despite executives knowing about the child sexual exploitation
problems on the platform they have not committed sufficient
resources to detect, remove, and prevent this harmful content.
This is a news story that I would like to ask be made part of
the record.\1\
---------------------------------------------------------------------------
\1\ The information referenced by Senator Portman appears in the
Appendix on page 163.
---------------------------------------------------------------------------
There are lots of issues here. One is you made the right
decision not to monetize this explosive conduct at this time,
which is really pursuing a pornography scheme as I see it. But
I wonder if you can give us a commitment today to halting this
program indefinitely so as to prevent the platform and bad
actors from making money off of child sexual material?
Mr. Sullivan. First may I say that we abhor CSAM, the
sharing of sexual material. I appreciate your work there. I
worked on this here and also at Meta, so I have been working on
this for years.
I made that decision to pause this idea. It was not a
product. It was a set of people had an idea that they thought
they might want to pursue. I said I want to look at all the
information here and learn about where we stand, what the risks
could be. I think this is how the system should work. We looked
at a product in its very early ideation and did the analysis
and got the perspectives, and said this is not appropriate for
us to be doing. So that is how the process went.
Senator Portman. OK. So you made a commitment today not to
pursue it?
Mr. Sullivan. We are not pursuing that.
Senator Portman. You have made a commitment not to pursue
it in the future?
Mr. Sullivan. We have no plans to pursue monetization of
adult content. That is correct.
Senator Portman. You have no plans to do it. Can you just
tell us you are not going to do it?
Mr. Sullivan. I am not planning to do it, no.
Senator Portman. You are not planning to.
Mr. Sullivan. I am not doing this.
Senator Portman. Just say you are not going to do it.
Mr. Sullivan. We are not planning to do it, no.
Senator Portman. Cannot get a ``planning'' out of there.
Not to, again, leave anybody out, Mr. Mohan, we have not
had a chance to talk yet. I want to ask you about something
that is important to this Committee, and I hope a way forward
in terms of legislating and regulating platforms. Your
platform's algorithms have been described as a ``black box,''
according to experts and researchers, meaning there is little
to no transparency in the algorithms. I am sure you have heard
that before.
The question is, is there a way to come up with a
transparency approach that makes sense as calls grow for
Congress to pass legislation? I like the idea of having much
better information than we have, getting behind the curtain and
getting into that black box.
That is why along with Senator Chris Coons I drafted this
legislation called the Platform Accountability and Transparency
Act. It would require the largest tech platforms to share data
with vetted, independent researchers and other investigators so
that we can know exactly what is happening with regard to the
privacy issues we talked about today, or content moderation,
product development, sexual exploitation issues, key industry
practices.
My question for you, Mr. Mohan, would you be supportive of
legislation like PATA to get at this need for transparency and
for us to be able to legislate with better information?
Mr. Mohan. Yes, Senator, I would be supportive of the
spirit behind that regulation. The reason why is because I
agree with you. I do think that transparency around our
practices, how we go about them, is an important thing. It is
the reason why we have invested so heavily in our quarterly
transparency report, which you may be familiar with.
It is also the reason why we, just a few weeks ago,
launched the YouTube research program, which is similar, in my
understanding, to what the act that you are referring to is
trying to get at, which is giving academic researchers access
to our raw data, obviously in a user privacy-sensitive way,
where they can derive metrics or derive insights of their own
based on that data. We have taken it a step further where we
will also provide technical support that these researchers
might need to get at the insights that they are looking for.
I am very bullish about that transparency program, and
based on the feedback that we hope to get from researchers look
forward to enhancing it in the future as well.
Senator Portman. We are following your YouTube research
program carefully. We are glad you created it. We want to see
what the results are and we want to be sure these are
independent individuals who will give actual information about,
what the algorithms are, again, what is in the black box so
that citizens can understand it better, and as legislators we
can legislate better. I think that is a positive step.
With regard to PATA, can I hear from the other members of
the panel how you feel about this legislation? We have shared
it with all of you. We hope to introduce it soon. Again, it
would be bipartisan, and it would be one that would, I hope,
give us a way forward as a first step. Mr. Cox.
Mr. Cox. Senator, thanks. I know our teams have been in
contact with yours on this. We are aligned that more
transparency about content on our platform is a good thing. It
is a good thing for the public. It is a good thing for the
company.
We also have an academic research program called Fort,
where we have designed privacy-protected ways of sharing
information with outside academics and researchers. We have
also released a widely viewed content report which helps folks
get access to which content is seen the most times on the
platform. We also publish quarterly community standards
enforcement report which gets into categories of content by
region and shows the work we are doing every day.
We are committed to working with you on this.
Senator Portman. Yes. We talked about regulations needs.
Ms. Pappas, yes or no?
Ms. Pappas. Senator, transparency builds trust. We were the
first open platform to open our own Transparency and
Accountability Center for that specific reason, so people could
take a look at our content moderation systems, recommendation
systems as well. Last month we announced that we will be
opening our API to researchers as well, so we would be happy to
support that legislation.
Senator Portman. OK. Thank you. Mr. Sullivan.
Mr. Sullivan. Yes. We have been publishing data to
researchers for years, and we are very open to anything that
improves transparency. Especially as AI moves forward, it going
to be very important.
Senator Portman. It is important. It is needed. Thank you,
Mr. Chairman. Thank you all.
Chairman Peters. Thank you, Ranking Member Portman.
Senator Johnson, you are recognized for your questions.
Senator Johnson. Thank you, Mr. Chairman. Mr. Cox, just a
quick little housekeeping here. Are you aware of a letter
Senator Grassley and I sent to Mr. Zuckerberg on August 29th?
We did get a reply on September 12th from Mr. Kevin Martin,
just saying they are going to respond. Are you aware of that
letter asking for information, contact between yourself, FBI,
Department of Justice, documents, names, that type of thing?
Mr. Cox. Senator, yes, I am aware of that letter and I know
the team is working on following up as quickly as they can.
Senator Johnson. You will commit to full response on that?
Mr. Cox. I know the team is committed to a response, yes.
Senator Johnson. OK. Let us put up my first chart\1\ here.
Back in November 2021, CDC Rochelle Walensky stated in front of
the Health Committee, ``We have the most robust, safe vaccine
safety system we have ever had in this country.''
---------------------------------------------------------------------------
\1\ The chart referenced by Senator Johnson appears in the Appendix
on page 192.
---------------------------------------------------------------------------
In October 2020, before the vaccine was approved, CDC's Dr.
Tom Shimabukuro stated in a web seminar, `` Vaccine Adverse
Event Reporting System (VAERS) is obviously something that Ms.
Walensky was talking about--``VAERS traditionally has provided
the initial data on the safety profile of new vaccines when
they are introduced. For COVID, vaccine reports will be
processed within one to five days. Depending on the seriousness
of the report, CDC and FDC received updated datasets daily, and
data-mining runs are planned to be conducted every one to two
weeks.''
This is an example of the timeliness and responsiveness of
VAERS, going back to H1N1. It kind of sounds like they are
really going to rely on VAERS. I remember part of that
discussion when they said, ``Listen, we are going to take
vaccine safety so seriously, if we get a report of a couple of
days of lost time because of an injury we are going to be
calling that individual up and we are going to be checking on
it.'' It really sounded like they had this all covered, right?
Let us see what they actually did. I produced this chart\2\
because I took VAERS and FAERS seriously, and I started
tracking this, and I started putting together this chart. I
want to quickly describe what this is. The first five lines,
the first five drugs, four of them are in the FDA Adverse Event
Reporting System (FAERS) system, the FDA Adverse Event safety
system: ivermectin, hydroxychloroquine, dexamethasone, and
Tylenol. You have the flu vaccines in there. That comes off the
VAERS system. You have remdesivir, which comes off of FAERS,
and COVID vaccine that comes off of the VAERS system.
---------------------------------------------------------------------------
\2\ The chart referenced by Senator Johnson appears in the Appendix
on page 193.
---------------------------------------------------------------------------
Now I can see why the government really did not like the
way I put this is their data. I did not make these numbers up.
This off the VAERS and the FAERS system. But for whatever
reason Twitter censored this chart.
Now just quick, and I will show you the current version,
the one you censored, it showed that ivermectin, on average,
over 26 years, on average, had 15 deaths reported on the FAERS
system. Hydroxychloroquine had 69 deaths. Flu vaccines had 77.
Dexamethasone had 618. Tylenol had over 1,000. Remdesivir,
since it was approved, had 1,612, and the vaccines had 21,000
deaths. OK, these are just the facts, and Twitter censored it.
Do you have any idea why?
Mr. Sullivan. Senator, I was not at the company at the
time, but what I can tell you is that we want robust discussion
on the platform, of any issue. A COVID misinformation policy
was developed that seemed to me--and again, I do not develop
it--but it seemed quite narrow to me.
Senator Johnson. You censored government information. Here
are the current numbers, by the way. Over 30,000 deaths
reported worldwide, 27 percent of those, by the way, have
occurred on Days zero, one, or two. You did not only censor
this chart, you censored, for example, radio shows that
interviewed me, talking about FDA-CDC data.
YouTube took down a video of this Committee's hearing, of
an eminently qualified critical care specialist who saved
thousands of lives treating people, using what seems to me
pretty safe drugs. After eight million views, YouTube pulled
that video down. What would be the justification for YouTube
pulling down a hearing of the U.S. Senate with a highly
qualified doctor just giving a second opinion on how to save
lives during COVID? Why would YouTube do that? On what
authority, whose authority, are you censoring that information
so the American public could not receive a second opinion, and
access drugs that might have saved their lives? Why would
YouTube do that?
Mr. Mohan. Senator, respectfully, as I was mentioning
earlier, we did not decide those policies on our own. We worked
with third-party health authorities in this country. That did
include the CDC or the FDA.
Senator Johnson. I will be sending you a letter, and I want
to know who those health authorities were, and I want to see
the communications between them. Will you commit to providing
me that information, for transparency's sake?
Mr. Mohan. Senator, I am happy to follow up on your request
on how we developed that policy.
Senator Johnson. In July 2021, talk about misinformation,
this should have been the 2021 lie of the year. President Biden
said, ``You are not going to get COVID if you have these
vaccines. If you are vaccinated you are not going to be
hospitalized, you are not going to be in an ICU unit, and you
are not going to die.'' That is the President of the United
States.
It just so happens we could not rely on the CDC and the FDA
because they were not honest, they were not transparent, they
were not giving us data, so we had to go to Public Health
England. This is a chart\1\ published from their Technical
Briefing Number 23, that covered the period from February 1 to
September 12, 2021. It shows 593 cases of mainly Delta, 2,542
deaths, 1,613 deaths occurred with the fully vaccinated.
---------------------------------------------------------------------------
\1\ The chart referenced by Senator Johnson appears in the Appendix
on page 195.
---------------------------------------------------------------------------
Obviously, this was published, and they were publishing
other similar information during that time period when
President Biden lied to the American public that this was a
pandemic event and vaxxed, and if you got vaccinated you are
not going to go to the hospital, you are not going to be in an
ICU unit, and you are not going to die. Well, 63.5 percent of
the people fully vaccinated were dying in England at the exact
same time.
Why did you not pull this? Have you ever labeled the
President of the United States' comment as misinformation? Have
you ever done that? Any of you? I will take that as a no.
Again, I just wonder, who are the authorities, who do you
think you are to censor information from eminently qualified
doctors who had the courage and compassion to treat COVID
patients when the National Institute of Health (NIH) guideline
was basically if you test positive for COVID, go home, be
afraid, isolate yourself, do not do anything until you are so
sick, we will send you to the hospital, we will give you
remdesivir, where we have 1,600 deaths so far, we will put you
on a vent, and we will watch you die.
You guys bear a fair amount of responsibility for hundreds
of thousands of people not being treated, and I would say
probably dying that did not have to die. I hope you are proud
of yourselves.
Chairman Peters. Senator Lankford. Here now in the second
round Senator Lankford, Senator Hawley, and then you.
Senator Lankford. I will give back part of my magic minute
here.
Chairman Peters. Yes, please do that.
Senator Lankford. I will go short on this. I do want to
follow through on a couple of things there on illegal activity.
You have all been very outspoken on dealing with sexual child
predators, with different issues, drug trafficking. Those were
all good things to be able to engage on.
But it is fascinating to me that the platforms have chosen
to say there are some illegal activities we are OK with, and,
in fact, we are going to facilitate. One of those is illegally
crossing our Southern Border. It is not hard for me to go to
YouTube, and I just type in ``how to cross the border
illegally'' and I get a video that says, ``How to illegally
cross the Mexico-U.S. border.'' It has 1.7 million views, and
it has been there for two years.
Yes, I watched it, and it showed where to be able to cross,
what highways to avoid, where the Border Patrol typically puts
up stations, how to be able to look for different aspects. In
detail, shows a video of here is how to illegally cross the
Mexico-U.S. border, and where to be able to cross, and how to
avoid border patrol. This has been up for two years, and it has
had 1.7 million views.
As I mentioned, on Facebook, Facebook has ads that I can
actually show you that are human smugglers placing ads in
Central America so people will know how to be able to connect
with them, to be able to travel through Mexico, to be able to
pay the cartels, which are a ruthless drug organization, to be
able to get in the United States.
My confusion on this is I do not understand why the
platforms look at illegally crossing the border as ``we are
going to look the other way'' when your user agreements say
``we do not promote illegal activity except for this one.''
Help me understand why that is different.
Mr. Mohan. Senator, I do not know about that specific
video. I am happy to follow up.
Senator Lankford. It is not just one. It is a bunch. That
is just the first of many.
Mr. Mohan. I am happy to follow up on those.
But just in general, we do have very clear policies where
content that encourages dangerous behavior, not just illegal
behavior but dangerous, harmful behavior is removed from our
platform. We have the Four R's approach that I described in my
opening testimony, where it not about removal of content but
also reduction of content, raising up authoritative sources. In
the context of people searching for that type of information,
making news stories from prominent mainstream news outlets
prominent. We do try to have a holistic approach to dealing
with this type of content on our platform.
We are not perfect. We continue to improve both our
policies as well as our enforcement. In this specific case I am
happy to follow up. But we do have very clear policies against
cartels, harmful criminal conspiracies, other types of
organizations where their type of activity is not allows on our
platform.
Senator Lankford. I would only say this particular video,
which, by the way, this one is in English, this particular
video even talks about how to be able to connect with a cartel
and how much the cost is going to be when you get to the
Southern Border.
Mr. Mohan. Senator, I am happy to follow up, but we do take
our enforcement----
Senator Lankford. I get it. This part is not being
enforced. That is what I am trying to say to you, is that I do
see all the platforms trying to deal with drug trafficking, but
human smuggling and illegally crossing the border is not being
enforced. I am not asking you to solve it today. I am raising
it as an issue to say somehow we treat cartels different than
terrorist organizations.
Cartels are transnational criminal organizations that are
making money off of moving people illegally into our country
and making money off of illegally moving drugs into our
country. I would like for our social media platforms to engage
with a criminal organization and with criminal activity
consistent to your own terms of service.
That is it. I yield back my time.
Chairman Peters. Thank you, Senator Lankford. Senator
Hawley.
Senator Hawley. Thank you, Mr. Chairman.
Mr. Cox, I know that Facebook has said in the past that it
is their position, as a private company, you are not subject to
the First Amendment. I assume that has not changed. Is that
right?
Mr. Cox. That is correct, Senator.
Senator Hawley. But the United States government is subject
to the First Amendment. I think we can probably all agree on.
Hopefully we can. Hopefully that is still true in this country.
Is it appropriate for Facebook to work with the United
States government to avoid the First Amendment, help the U.S.
Government avoid the First Amendment?
Mr. Cox. Senator, we do think it is sometimes appropriate
to be in contact with government and with government
organizations.
Senator Hawley. To help them avoid the First Amendment?
Mr. Cox. Senator, I am not sure what specifically you are
referring to.
Senator Hawley. Let me ask you this. Do you think it is
appropriate to work with the United States government to target
private individual speech that is constitutionally protected?
Mr. Cox. Senator, I am not aware of that.
Senator Hawley. Let me educate you. On July 16, 2021, an
employee at Facebook wrote to the Department of Health and
Human Services, saying, ``I know our teams met today to better
understand the scope of what the White House expects from us on
misinformation going forward.''
On July 23, 2021, a Facebook employee thanked HHS, quote
``for taking the time to meet earlier today, and wanted to make
sure you saw the steps we just took this past week to adjust
policies on what we are removing with respect to
misinformation. This included''--and I am still quoting--
``increasing the strength of our demotions for COVID and
vaccine-related content.''
On April 7, 2021, a Facebook employee thanked the CDC for
responding to misinformation queries, and I quote, ``We will
get moving now to be able to remove all but that one claim as
soon as the announcement and authorization happens.''
On July 28th of this year, a Facebook employee reached out
to CDC about, ``doing a monthly misinfo/debunking meeting.''
The CDC responded, ``Yes, we would love to do that.'' I am sure
they would.
On July 20, 2021, Clark Humphrey at the White House, who
was digital director of the COVID-19 response team, emailed
David Sumner at your company, among others, asking, ``Any way
we can get this pulled down,'' and cited a specific Instagram
account. Within 46 seconds, your company replied and said,
``Yep. On it.'' That sounds like what, in the law, we call a
pattern and practice of meeting, coordinating, and colluding
with the United States government to target particular speech
that no one in any of these emails alleges is incitement, which
would not be constitutionally protected, no one in any of these
emails alleges it directly encourages violence, which would not
be constitutionally protected.
It appears to all be constitutionally protected speech on,
I might add, very politically sensitive topics, that Facebook
is directly working with the U.S. Government to target and
remove. Is that your company policy to do this kind of thing?
Mr. Cox. Senator, we were quite public about our
cooperation with health organizations during the unprecedented
time of COVID. We knew that people expected and wanted accurate
information on our platform. We had conversations with CDC,
with the World Health Organization (WHO), and with other public
health organizations, not just in the United States but abroad,
in order to understand how to help make sure that folks were
not getting information that could cause them any harm.
Senator Hawley. Fair enough. You are saying that this was,
in fact, company policy to have these kinds of meetings with
HHS, with the CDC, with the White House directly, that you did
engage in this behavior, and you think that it was entirely
fine. Is that your testimony?
Mr. Cox. Senator, I do believe it is appropriate for
companies like ours to be in consultation with public health
organizations and with government.
Senator Hawley. You can confirm that things like taking
down a private Instagram account and adjusting your policies at
the behest of the White House, and putting into place
misinformation policies at the behest of CDC, that those
things, you think, are appropriate, that this was company
policy to do so. Is that fair to say?
Mr. Cox. Senator, I am not familiar with the Instagram
account specifically that you are referencing, but we do know
that people expected and hoped from the platforms that we would
help them get accurate information about COVID during the
unprecedented time, especially at the beginning.
Senator Hawley. Is there not a difference between you, as a
platform, putting forward information, and censoring your users
at the behest of the White House, the Administration more
broadly, and the CDC? Is there not a distinction there?
Mr. Cox. We specifically wanted to work with public health
experts to understand the relationship between information and
behavior, and so we did consult with the CDC, the World Health
Organization, and others to understand how the platform
policies we built were affecting public health.
Senator Hawley. You did not just consult them to understand
how they affected public health, you actually censored on their
behalf. I mean, you took these emails--I am just quoting from a
sample of them--which, by the way, have been disclosed in
litigation--these emails show that you took censorship steps.
You took down accounts. You planned misinformation policies.
You adjusted your policies at the behest of the United States
government. That is not just some theoretical thing. That is
actually targeting your users' speech.
I appreciate your forthrightness, by the way.
But you think that is fine, and that was your policy.
Mr. Cox. Senator, we have been public about our policies,
on COVID misinformation specifically, as well as on
misinformation generally.
Senator Hawley. You are not concerned about any of this.
Nothing that I just read to you, you are not concerned about it
at all.
Mr. Cox. Respectfully, Senator, I think the balance of how
to protect free expression as well as public safety is a
difficult issue, but it one we are committed to working with
outside experts and publishing our work.
Senator Hawley. I appreciate you being so forthright. As I
said, this is actually from litigation between the State of
Missouri and the State of Louisiana and the Federal Government.
I anticipate that your remarks under oath today are going to be
very interesting and helpful to that litigation.
I will just say this. My view is that the United States
government is bound by the First Amendment. They cannot
encourage or coerce or incite or collude with a private party
to get around the First Amendment, that you have just said to
me today that that is basically what they did, that you
coordinated with them repeatedly, over a pattern of months and
years, to adjust and target your speech policies for protected
speech at the behest of the United States government.
I have to tell you, I have a big problem with that, and I
think all your users should too.
Thank you, Mr. Chairman.
Chairman Peters. Thank you, Senator.
To our panelists, it is at 5 p.m. I know there was a
suggestion for a break. We are right down to the end here.
Rather than break and come back and keep you here longer we are
going to power right through it with Senator Scott, and then I
will wrap it up after that. Senator Scott.
OPENING STATEMENT OF SENATOR SCOTT
Senator Scott. All right. Thank you, Chair Peters. Thanks
to each of you for being here.
It is critical for employees and officers of the FBI and
DOJ to continue to have a mechanism for reporting concerns of
fraud, waste, and abuse within their respective agencies to
Congress without fear of reprisal from DOJ/FBI leadership.
Whistleblowers from within the DOJ/FBI have come forward
with concerns about the Department of Justice's alleged
political bias in the FBI's raid on the former President's home
in Florida last month. FBI agents have reported similar
concerns to individual Senate offices as well. We all need to
ensure safeguards are in place so Attorney General Garland does
not retaliate against or intimidate FBI agents and DOJ
employees who come to Congress as whistleblowers.
Mark Zuckerberg recently disclosed that Facebook's
restriction of a story about Hunter Biden during the 2020
election was based on the FBI's, ``misinformation warnings.''
Additionally, emails and internal communications obtained by
the journalist Alex Berenson, in his lawsuit against Twitter,
have shown his removal from the social media platform was a
result of pressure from Biden White House officials to silence
his criticism of the Administration's COVID-19 policies.
These instances and several others show a clear and
alarming pattern of speech suppression carried out at the
direction of agencies and officials in the Federal Government.
In other words, the Federal Government used private businesses
to violate the First Amendment rights of our citizens. This
also confirms that the Federal Government used officials at the
FBI to interfere in the 2020 election by manipulating the
normal flow of public discourse and information-sharing with
false warnings about foreign interference and disinformation.
I am going to ask you a couple of questions, if you can
just show by hands yes or no. By a show of hands, how many of
you and your companies have been contacted by a Federal agency,
an agency official, or a member of the Biden White House with a
request to remove, censor, or restrict access to a post or an
individual user on your platform? If you have, would you raise
your hand and say yes.
Mr. Sullivan. I am not aware of it.
Senator Scott. So Meta, YouTube, TikTok, and Twitter all
said never.
Mr. Cox. Not to my knowledge, Senator.
Senator Scott. You would know, would you not?
Mr. Cox. I have not been in conversations with the FBI,
Senator.
Senator Scott. OK. So no conversations. OK.
By a show of hands, how many of you and your companies have
felt pressure to remove, censor, or restrict access to a post
or individual user on your platform based on that contact with
a Federal agency or official?
Is the answer no from all four of you? So all four of you
say no.
Mr. Cox. Not to my knowledge, Senator.
Senator Scott. OK. By a show of hands, how many of you and
your companies have received a misinformation warning issued by
the FBI? Every one of you is saying no?
Mr. Cox. Senator, I know that we have received warnings
from the FBI and other experts about electoral misinformation,
in general, and foreign interference, in general.
Senator Scott. By a show of hands--but the rest of you have
said no.
Mr. Mohan. Senator, we receive information from every
administration about things like foreign interference and
election results, et cetera. But one thing is very clear. We
enforce our guidelines based on our community guidelines, so we
are the ones making decisions about the content that is
removed, based on the guidelines that we publish transparently,
not based on what a particular administration asks us to do or
not do.
Senator Scott. So your answer is you have never received a
misinformation warning issued by the FBI?
Mr. Mohan. Senator, no. What I am saying is that we do
receive information from the FBI in terms of imminent threats,
foreign actors trying to interfere with our free and fair
elections here in the United States, and we take that into
account in terms of the enforcement of our policies. But those
decisions about the enforcement of our policies are made solely
based on our community guidelines that we publish on our
website.
Senator Scott. Mr. Cox, what I said about Mark Zuckerberg
disclosed the FBI's restriction of a story about Hunter Biden
during the 2020 election was based on the FBI's misinformation
warning, is that untrue?
Mr. Cox. Senator, I was not in conversations with the FBI
so I cannot speak to exactly what the conversation was. What I
do know is that we were in contact with a number of
organizations who warned us in the time leading up to the 2020
election, to be on the lookout for foreign interference in
elections, and that is an issue that we take incredibly
seriously.
Senator Scott. But you should know if Mark Zuckerberg,
right, would you not know----
Mr. Cox. Sorry, Senator.
Senator Scott [continuing]. Would you not know if Mark
Zuckerberg--he said that Facebook made that decision. You would
know that, right?
Mr. Cox. Senator, I do know that, if you are talking about
the New York Post story, I do know that consistent with our
policies we made the decision to submit that story to be
reviewed by independent fact-checkers. It was never removed
from our service, and we never blocked anybody from sharing
that story.
Senator Scott. I know people in Florida that were kicked
off by putting a story up.
Last one. By a show of hands, how many of you and your
companies have felt pressured to act upon an FBI misinformation
warning you received, such as by removing, censoring, or
restricting access to a particular user or post as a subject of
that misinformation warning?
So the answer is no for all of you.
Mr. Cox. Not that I am aware of.
Senator Scott. No for all of you. So every one of you said
no. All right.
Mr. Cox, in 2013, DOJ shut down Silk Road, the illegal
online marketplace on the dark web which featured over 13,000
illegal drug postings. In comparison, according to the 2020
Facebook transparency report, Facebook found 5.9 million
illegal drug sale postings on Facebook and Instagram. That is
453 times more drug postings than Silk Road. If you found
almost six million postings, how many posts are you not
finding? If that is true, if Silk Road got shut down, what in
the living daylights are you guys still doing in business?
Mr. Cox. Senator, we release quarterly reports on the
specific answers to your question around each category of bad
content on the platform. We do not believe the sale of illegal
drugs has a place on any of our platforms. We work hard to
fight against that. We publish quarterly the updates on exactly
how many pieces of content we take down, as well as how much we
are about to take down proactively.
Senator Scott. All right. I have one more question, and
this has to do with what has happened with my sheriffs. Do you
collect stats on the average turnaround time for responding and
resolving customer complaints like hijacked accounts or
products that violate your terms of service?
Mr. Sullivan, you say yes. No one else does?
Mr. Sullivan. Yes. We have a goal of meeting a service
level agreement to turn those around as quickly as possible.
Mr. Mohan. Senator, we do look at how quickly we respond to
requests from our creators who upload content to our platform
as well as viewers, and we are constantly looking to continue
to improve our processes around that sort of request handling.
Senator Scott. TikTok?
Ms. Pappas. We do as well, yes.
Senator Scott. And you do it also?
Mr. Sullivan. Are you asking whether we look at turnaround
times?
Senator Scott. Yes.
Mr. Sullivan. Yes we do, Senator.
Senator Scott. OK. Do you collect stats on the average
turnaround time for responding to subpoenas issued by law
enforcement agencies?
Is that a yes for everybody?
Mr. Mohan. Yes, Senator. We have a group that works to
respond to subpoenas, evaluate them, and respond 24/7.
Senator Scott. OK. I wrote to all of your companies and I
asked, and none of you responded. I do not know if you all
realize that. Every one of you, I wrote and asked for
information. None of you responded to either of those
questions.\1\
---------------------------------------------------------------------------
\1\ The Google letter in response to Senator Scott appears in the
Appendix on page 197.
---------------------------------------------------------------------------
Let me just tell you what sheriffs are saying. One of the
Florida Sheriff's Departments mentions, ``There is no point of
contact to send subpoenas. They are slow to respond. There is
no sense of urgency on how they respond to something that is
even time sensitive.''
I can tell you I have talked to sheriffs all around
Florida--is that time-sensitive information that would impact a
law enforcement investigation or a crime, you guys do not
respond. How do you respond to that?
Mr. Sullivan. I work in the product organization. We would
not be the ones to receive that, but I can have our team get
back to you.
Senator Scott. Anybody else?
Mr. Mohan. Senator, I am happy to follow up. We do have a
team that responds to those requests. We balance the needs of
law enforcement as well as our user privacy when we are
responding to those, as you would imagine. We do take into
account time sensitivity in terms of trying to respond to those
requests. I will ask our team to follow up with you as well.
Senator Scott. The questions I asked before that I sent you
all, that none of you responded to, you will respond to?
Mr. Cox. Senator, I would be happy to have the team follow
up. We take law enforcement requests very seriously.
Senator Scott. OK. Thank you. Thank you, Chair.
Chairman Peters. Thank you, Senator Scott.
I have a couple of follow-up questions and we will wrap up
and you will be on your way, and again, thank you so much for
taking so much time. You have been in the seat a long time. I
have been here with you a long time. I am ready to get out of
the seat as well, with you.
You have all spoken about how essential trust and safety
is, how it is part of your culture and something that you are
focused on. But I want to reference a report that was
commissioned by a Twitter whistleblower in the spring of 2021,
focused on its site's integrity team and misinformation. It
found that, ``Project managers are incentivized to ship
products as quickly as possible, and thus are willing to accept
security risk.''
Mr. Sullivan, as head of Consumer Products at Twitter,
would you agree with this finding in that report, commissioned
by Twitter?
Mr. Sullivan. The dates of that report would have been
before I was in the role, but what I can tell you is that I
have been in multiple product reviews where I push hard, and
our other leaders push very hard, and work with our teams to
strike a balance of safety in all that we do. I cannot speak to
that report but what I can tell you is how we operate now, and
that is how we operate now.
Chairman Peters. Obviously this report has a completely
different conclusion than you have. I just have to ask you, how
can you say that trust and safety are important to your
development progress when Twitter launched its Spaces product,
despite your predecessor publicly stating that Twitter would
not be able to moderate all of its Spaces?
It is my understanding that since the launch it is
documented that Spaces has been used by white supremacists as
well as ISIS to spread misinformation, as shown in this poster.
In fact, the internal report shown on this poster basically
says, if I can read it here, ``We did not prioritize
identifying and mitigating against health and safety risks
before launching Spaces.'' Do you disagree with this
characterization? I have heard you say on the record, ``We
never, ever send anything out that we have concerns about.''
This is obviously very different.
Mr. Sullivan. Yes, I understand what you are saying. Since
I have started in my role, we have been looking at health and
safety across the board and working to improve it.
In Spaces, for example, we have been continuing to beef up
all of our reporting, our automatic detection, our language
support. We are working very hard to further improve the health
of Spaces. I think that is just one example of many that I
could give you for how we are operating now.
Chairman Peters. After the fact. After some of these things
are released.
Mr. Cox, you also have talked quite a bit about trust and
safety as being central to Meta's development process. My
question to you is, why, after several years of warnings by
external organizations such as the Tech Transparency Project,
does Facebook continue to automatically generate home pages for
white supremacists and other extremist groups and terrorist
such as ISIS, as shown in this poster right here for the Aryan
Brotherhood, a page that has been created? I guess it was taken
down just recently, but it was on Facebook for 12 years, for 12
years the Aryan Brotherhood.
Does not this feature allow extremist groups to basically
recruit members more easily because you are putting this up?
Mr. Cox. Senator, we believe there is no place for
terrorism or violence-inciting networks, for militarized social
movements. We believe there is no place for these on our
platform. We use automated tools to find and take them down as
well as teams of experts dedicated to these specific problems.
Chairman Peters. Your automated tools and teams, they were
successful after 12 years. They were able to bring it down
after 12 years. Do you think that is an acceptable performance?
Mr. Cox. Senator, I know that for 97 percent of terrorist
content we are able to get to it before anybody reports it, and
also that we have been able to improve that number, quarter
over quarter, and I will continue to make sure that we aspire
to.
Chairman Peters. But certainly any content that gets
through and is disseminated very broadly can have catastrophic
consequences and violent actions, and particularly groups like
this where they have pages that are being put up by your
company that are there for 12 years. I hope you would agree
that that is unacceptable.
Mr. Cox. Senator, respectfully, we would not have put this
page up ourselves, but we do work hard to make sure that
extremist and terrorist networks are not allowed----
Chairman Peters. Yes, this is auto-generated. This was an
auto-generated page.
Mr. Cox. Senator, I have not seen this specific example.
Chairman Peters. I would love to have you comment on this.
If you could look at this example, and if we could have your
written comments on it would appreciate that afterwards.
My final question, and then we are going to let you go,
seriously. When your product teams are testing new products or
features I know that you track engagement and growth, a pretty
fundamental part of the work that you do.
My question to you, and this is to each of you, do you
consistently measure the impacts of these new features on
societal harms like misinformation, disinformation, hate
speech, and terrorism? If you could just give me a yes-or-no
answer. We will start with you, Mr. Sullivan, and work that
way. Yes or no, please.
Mr. Sullivan. Yes, depending on the feature and which of
those harms might apply, we go deep into those and analyze
those, yes.
Chairman Peters. So it is yes. If you could go a little
deeper I am going to ask you then, how do you characterize and
measure this data, and which metrics are used? If you can be
very specific, it is very helpful that we have that
information. I will ask all three of you to do the same,
please.
Mr. Sullivan. I will give you one topline metric we use for
many of these, which is what we call ``harmful impressions.''
So 0.1 percent of tweets that turn out to be violative, we want
to limit the number of people that see them. It is how many
people may have seen something before we identified it as
violative. Those are some of the metrics that would be
important to combat this harm on the platform. Thank you,
Senator.
Chairman Peters. Thank you. Ms. Pappas.
Ms. Pappas. Similarly the way we measure it is looking at
community guidelines violation rate. Essentially we do a sample
size that is view-weighted of our corpus of videos, and then we
are looking at was there any violative content, and then we are
looking at how do we minimize that exposure and drive it down
to zero.
As I had mentioned earlier, we look at that at a per-policy
basis, so things like hate speech, violent extremism, mis-and
disinformation, and we are able to measure our improvement of
that, quarter over quarter, week over week. We look at those
metrics and reports, and doing so with regard to our features,
like our For You feed.
Chairman Peters. You do this before every launch, this is
the study you will do?
Ms. Pappas. We would not have the metrics before launch,
but in terms of setting our baselines and knowing----
Chairman Peters. Right. Mr. Mohan.
Mr. Mohan. Senator, I can say very clearly that our
responsibility as a global platform comes before any of our
growth metrics around engagement, revenue, et cetera. It is the
top line of the company, and we are constantly reviewing our
products.
Chairman Peters. How do you characterize the data?
Mr. Mohan. When we are reviewing our products on a constant
basis one of the core metrics that I look at, and that the rest
of the company looks at, is something called our ``violative
view rate (VVR).'' That is the metric that we have also started
to publish on a quarterly basis so that you have access to it.
In fact, our most recent transparency report was just published
a couple of weeks ago. The violative view rate is basically a
metric that calculates how much content is up on our platform
that would have violated any of our policies, across hate
speech, harassment, et cetera. That number is something on the
order of 9 to 11 impressions out of 10,000. It is a small
number that we aim every single quarter to continue to drive
down.
Chairman Peters. I want to be clear. You do this with your
AB testing (split testing). This is testing that you do before
you launch a product?
Mr. Mohan. We measure this metric on a constant basis.
Chairman Peters. But do you do it before you launch a
project?
Mr. Mohan. We would not have the metric----
Chairman Peters. You do testing. You do not launch without
doing some testing. You do AB testing. I have not heard about
AB testing here from you today. You do not test the product
before you launch? You just say, ``Hey, let's launch and see
what happens''? I do not think you do that. Is that what you
do? You launch products without testing whether or not it makes
sense?
Mr. Mohan. Senator, I did not say that.
Chairman Peters. OK.
Mr. Mohan. We test our products extensively before
launching, in terms of usability of the products, but also the
trust and safety and the impact those products will have on our
users.
Chairman Peters. OK. You do those tests during the test.
Mr. Mohan. Correct.
Chairman Peters. Ms. Pappas, you said afterwards. Do you do
that AB testing before you launch, and do you also test safety
and trust, because you are testing your product before you send
it--unless you do not test it before you send it out to the
world.
Ms. Pappas. We do do testing before launch, and we will
delay products, or rather features, if they do not meet our
safety standards.
Chairman Peters. Mr. Cox.
Mr. Cox. Senator, to your question, the primary way we
measure and understand is prevalence, and we publish reports on
the specific categories of content that violates our policies
as well as regions around the world.
Chairman Peters. You do that in your AB testing before you
launch?
Mr. Cox. Senator, for many of those metrics you need a
specific study in order to understand that metric, but we look
at lots of other metrics associated----
Chairman Peters. Does that mean you do not do it before you
launch?
Mr. Cox. What we would do before the launch of any product
where we had any reason to be concerned about safety is put it
through a review with our integrity teams, whose job is to
understand safety concerns. We would not launch a product if we
believed that there was a safety issue.
Chairman Peters. You have referenced several times the
statistic that hate speech on your platform represents 0.02
percent of all views. Is that accurate?
Mr. Cox. That is correct.
Chairman Peters. Certainly that sounds like a small number.
I will appreciate that. But you also have a lot of views. You
are a massive platform. I am concerned that this could mask the
total amount of hate speech that could be out there and viewed
by an awful lot of folks. My question is, what is the total
number of views that hate speech actually gets on your
platform, not a percentage, but how many views, last year, for
example, or last month, yesterday, whatever you may have? Do
you have those numbers?
Mr. Cox. I do not have those numbers right now, Senator,
but I would be happy to have our teams follow up.
Chairman Peters. Would you provide those numbers to the
Committee as to the total number of views, not as a percentage?
Mr. Cox. Yes, I would be happy----
Chairman Peters. You can do the math. You have massive
amount of views on your platform.
Mr. Cox. Yes, Senator. I would be happy to have our teams
follow up on that.
Chairman Peters. I appreciate it. Thank you.
Thank you. I want to thank again our witnesses for joining
us today. I am certainly grateful for your contributions to
what is a very serious and a very important discussion, and I
want to certainly thank Ranking Member Portman for holding this
hearing with me here today.
I think today's hearing shed some new light on some serious
problems of rising domestic extremism and violence and its
relationship to amplified content on platforms.
We heard from our first panel earlier today about how user
engagement and revenue generation are the primary incentives
that drive product development and decisionmaking at your
companies, and that the overall goals of growth and profit are
always prioritized over the safety of users. That tradeoff,
revenue over safety, has contributed to, unfortunately, some
real-world harms, from horrific attacks and acts of violence
motivated by extreme ideologies to our fundamental democratic
process also being challenged.
I will be honest. I am frustrated that the Chief Product
Officers, who all of you have a prominent seat at the table
when these business decisions are made, were not more prepared
to speak to specifics about your product development process,
even when you were specifically asked if you would bring
specific numbers to us here today, and that your companies
continue to really avoid sharing some very important
information with us. We have been working on this for quite
some time and continue to be frustrated at the slow response,
or the no response that we receive from you.
The testimony we heard today from both experts and from
former executives as well as from the four of you have made
clear the important work of the current trust and safety teams.
It is simply not enough to address the problem. This problem
continues to be with us today.
Although we head plenty of testimony about your companies'
content moderation policies, what content gets removed and why
and even how much you spend on safety measures, it is clear
that those actions cannot effectively address this problem as
long as the product development process and the revenue-based
incentives do not change to make safety a higher priority in
those structures.
We need to continue this important conversation. This will
be the first of, I am sure, many conversations, and discuss
possibly regulatory measures and changes to the incentive
structures within your companies to build better practices, to
limit the spread of harmful and extreme content before it is
actually spread to users. Certainly we appreciate actions that
are taken after the fact, but at that point much harm could
already be released out into society with potentially
catastrophic consequences. We all want to be ahead of the
problem, not reacting to a problem that exists already in our
society.
As Chairman of this Committee I will continue to work
alongside Ranking Member Portman and Members of the Committee
to find effective solutions to this growing homeland security
threat. I certainly hope that each and every one of you will be
part of that process to find that solution. We all need to be
working together on this. It is very clear, the more we talk
about this issue, the more we realize how complex it is, and it
is going to take all of us putting our heads together and
figuring out a path forward.
The record for this hearing will remain open for 15 days,
until 5 p.m. on September 29, 2022, for the submission of
statements and questions for the record.
With that, this hearing is now adjourned.
[Whereupon, at 5:25 p.m., the hearing was adjourned.]
A P P E N D I X
----------
[GRAPHIC(S) NOT AVAILABLE IN TIFF FORMAT]