[House Hearing, 118 Congress]
[From the U.S. Government Publishing Office]
WHITE HOUSE OVERREACH ON AI
=======================================================================
HEARING
BEFORE THE
SUBCOMMITTEE ON CYBERSECURITY, INFORMATION
TECHNOLOGY, AND GOVERNMENT INNOVATION
of the
COMMITTEE ON OVERSIGHT
AND ACCOUNTABILITY
U.S. HOUSE OF REPRESENTATIVES
ONE HUNDRED EIGHTEENTH CONGRESS
SECOND SESSION
__________
MARCH 21, 2024
__________
Serial No. 118-96
__________
Printed for the use of the Committee on Oversight and Accountability
[GRAPHIC NOT AVAILABLE IN TIFF FORMAT]
Available on: govinfo.gov
oversight.house.gov or
docs.house.gov
__________
U.S. GOVERNMENT PUBLISHING OFFICE
55-220 PDF WASHINGTON : 2024
-----------------------------------------------------------------------------------
COMMITTEE ON OVERSIGHT AND ACCOUNTABILITY
JAMES COMER, Kentucky, Chairman
Jim Jordan, Ohio Jamie Raskin, Maryland, Ranking
Mike Turner, Ohio Minority Member
Paul Gosar, Arizona Eleanor Holmes Norton, District of
Virginia Foxx, North Carolina Columbia
Glenn Grothman, Wisconsin Stephen F. Lynch, Massachusetts
Michael Cloud, Texas Gerald E. Connolly, Virginia
Gary Palmer, Alabama Raja Krishnamoorthi, Illinois
Clay Higgins, Louisiana Ro Khanna, California
Pete Sessions, Texas Kweisi Mfume, Maryland
Andy Biggs, Arizona Alexandria Ocasio-Cortez, New York
Nancy Mace, South Carolina Katie Porter, California
Jake LaTurner, Kansas Cori Bush, Missouri
Pat Fallon, Texas Shontel Brown, Ohio
Byron Donalds, Florida Melanie Stansbury, New Mexico
Scott Perry, Pennsylvania Robert Garcia, California
William Timmons, South Carolina Maxwell Frost, Florida
Tim Burchett, Tennessee Summer Lee, Pennsylvania
Marjorie Taylor Greene, Georgia Greg Casar, Texas
Lisa McClain, Michigan Jasmine Crockett, Texas
Lauren Boebert, Colorado Dan Goldman, New York
Russell Fry, South Carolina Jared Moskowitz, Florida
Anna Paulina Luna, Florida Rashida Tlaib, Michigan
Nick Langworthy, New York Ayanna Pressley, Massachesetts
Eric Burlison, Missouri
Mike Waltz, Florida
------
Mark Marin, Staff Director
Jessica Donlon, Deputy Staff Director and General Counsel
Peter Warren, Senior Advisor
Lauren Lombardo, Deputy Policy Director
Raj Bharwani, Senior Professional Staff Member
Mallory Cogar, Deputy Director of Operations and Chief Clerk
Contact Number: 202-225-5074
Julie Tagen, Minority Staff Director
Contact Number: 202-225-5051
------
Subcommittee on Cybersecurity, Information Technology, and Government
Innovation
Nancy Mace, South Carolina, Chairwoman
William Timmons, South Carolina Gerald E. Connolly, Virginia
Tim Burchett, Tennessee Ranking Minority Member
Marjorie Taylor Greene, Georgia Ro Khanna, California
Anna Paulina Luna, Florida Stephen F. Lynch, Massachusetts
Nick Langworthy, New York Kweisi Mfume, Maryland
Eric Burlison, Missouri Jared Moskowitz, Florida
Vacancy Ayanna Pressley, Massachesetts
Vacancy Vacancy
C O N T E N T S
----------
Page
Hearing held on March 21, 2024................................... 1
Witnesses
----------
Mr. Adam Thierer, Resident Senior Fellow, Technology &
Innovation, R Street Institute
Oral Statement................................................... 5
Ms. Jennifer Huddleston, Technology Policy Research Fellow, Cato
Institute
Oral Statement................................................... 7
Mr. Neil Chilson, Head of AI Policy, The Abundance Institute
Oral Statement................................................... 8
Dr. Nicol Turner Lee (minority witness), Senior Fellow,
Governance Studies and Director, Center for Technology
Innovation, Brookings Institution
Oral Statement................................................... 10
Written opening statements and statements for the witnesses are
available on the U.S. House of Representatives Document
Repository at: docs.house.gov.
Index of Documents
----------
* No additional documents were submitted for this hearing.
WHITE HOUSE OVERREACH ON AI
----------
Thursday, March 21, 2024
U.S. House of Representatives
Committee on Oversight and Accountability
Subcommittee on Cybersecurity, Information Technology, and Government
Innovation
Washington, D.C.
The Subcommittee met, pursuant to notice, at 10:04 a.m., in
room 2154, Rayburn House Office Building, Hon. Nancy Mace
[Chairwoman of the Subcommittee] presiding.
Present: Representatives Mace, Timmons, Burchett, Burlison,
Connolly, Lynch, and Pressley.
Ms. Mace. Good morning, everyone. The Subcommittee on
Cybersecurity, Information Technology, and Government
Innovation will come to order. Welcome, everyone.
Without objection, the Chair may declare a recess at any
time.
I would like to now ask unanimous consent for
Representative Don Beyer from Virginia to be waived on to the
Subcommittee for today's hearing for the purpose of asking
questions. So, without objection, so ordered.
I will now recognize myself for the purpose of making an
opening statement.
Good morning. Turn that frown upside down. It is going to
be OK. Last October 30, the White House released a monumentally
lengthy executive order on artificial intelligence, and the EO
is not just long, it is broad. It corrals dozens of Federal
agencies into a massive posse that is to go out and ride herd
on every aspect of this emerging technology. But why the
stampede? We are only just starting to grasp how AI can help
and also harm humanity. That is why Congress is moving
cautiously in this space.
We already have a plethora of laws in which AI uses are
subject, ranging from anti-discrimination to consumer
protection statutes, and unnecessary new laws could stifle AI
innovation, slowing the arrival of life-enhancing and
lifesaving breakthroughs. Not to mention, we do not want China
on our heels, and we do not want to stifle innovation for the
private sector or for our government agencies, especially in
defense. That is why Congress is proceeding with a measured
first do no harm approach. Where AI applications are not
captured under an existing law, we need to close these
loopholes. That is why I introduced a bill recently to ensure
the distribution of nonconsensual pornography is not immune
from criminal prosecution just because it has been altered via
an AI deepfake process.
But Congress wisely has not authorized the Administration
to go out and regulate AI differently than other technologies.
But this executive order does so anyway and invokes the
emergency powers of the Defense Production Act, or DPA, to
require AI developers to notify the government if they are even
considering developing new AI systems. It also mandates they
regularly hand over highly sensitive proprietary data, like
testing results, to the Commerce Department. What does this
have to do with defense production?
The DPA gives the President extraordinary powers to ensure
the supply of critical goods in time of war or national
emergency, but we are not at war today, and if artificial
intelligence is an emergency, it is not a temporary one. AI is
not going to go away anytime soon, so the new executive powers
this EO asserts have no logical sunset.
The bottom line is that this use of the DPA appears to be
executive overreach and, quite frankly, illegal. That is the
view of the attorneys general of 20 states, including my own
home state of South Carolina. These AGs last month wrote a
letter to the Commerce Secretary that argued the reporting
regime in the EO lacks legal authority because the DPA allows
for the Federal Government to promote and prioritize
production, not to gatekeep and regulate emerging technologies.
I want to thank my Attorney General, Alan Wilson, for stepping
up to the plate and being a part of this letter.
What is more, the gatekeeping in this EO seems more likely
to harm than help our national defense. What is the biggest
national security concern around AI? It is the risk that we
relinquish our current lead in AI to China, and that could have
catastrophic implications for our military preparedness. But
requiring potential new AI developers to share critical data
about their most valuable assets with the government could
scare away would-be innovators and impede more ChatGPT-type
breakthroughs. Not only that, but they might take their
technology elsewhere, and we want the best AI developers, the
best AI programmers, the best AI tech right here in the United
States.
Also, how will the government protect that data it gets
from AI firms? The EO could risk national security by mandating
the creation of what Brookings Institution fellow, John
Villasenor, calls a target list for any geopolitical adversary
that might want to engage in cyber espionage or launch a large-
scale cyberattack on U.S. AI computing infrastructure. He notes
the government's poor record of preventing the exfiltration of
data from Federal computer systems by malicious actors,
including foreign enemies.
To be clear, the government does need to be proactive with
respect to artificial intelligence. The executive branch needs
to harness AI to strengthen national defense, bolster homeland
security, and improve the administration of benefit programs.
That is why I am glad that the EO contains numerous provisions
to strengthen the government's own AI workforce and to enhance
the government's ability to contract with private sector AI
providers. But as the rubber hits the road on this EO, with the
implementation deadlines already having begun to kick in, I
look forward to hearing from our panelists today about where
they believe it exceeds the President's legitimate authority
and where it could impede American innovation.
With that, I will now yield to Ranking Member Connolly for
his opening statement.
Mr. Connolly. Thank you. AI helps doctors optimize medical
treatments, scientists predict potential natural disasters, and
Federal workers manage water supplies, and yet we know very
well every scientific and technological advancement comes with
its own risks. The Biden-Harris executive order on AI, the
first of its kind ever, elegantly balances innovation with
equity and potential with pragmatism. AI has already
demonstrated massive and consequential effects on workforces
and economies around the globe. The Biden-Harris executive
order sets America on a path to lead the world in ethical,
equitable, and transparent use of AI.
As we enter the second hearing exploring the Biden-Harris
EO, it is good to remember what our previous witnesses
testified unanimously. One Republican witness, ``The AI
executive order and OMB memo are important steps that focused
on AI safety, investment, talent, and leadership, and are
critical for America to lead in AI innovation and governance,
but the executive branch cannot achieve this goal fully without
Congress.'' Another Republican Majority witness: ``In sum, the
AI executive order and OMB memo have taken a big first step,
but it is only one step in a longer journey. Congress must now
kick in.'' Another: ``The executive order recognizes the
importance of the AI Risk Management Framework developed by the
National Institute of Standards and Technology. We encourage
the Administration to ensure that framework anchors the
government's risk management efforts.'' These quotes all come
from Majority-picked witnesses and agree that the Biden-Harris
executive order promotes safe and responsible AI and puts us on
the right path for a strong AI future in the United States, but
they also recognize Congress has more work to do.
Just last week, we had a hearing on deepfakes and learned
how some AI training data sets contain known images of child
sex abuse materials. As a result, some AI models trained off
this data can further perpetuate the creation of additional
terrible and shocking content. This example is just one way AI
technology can cause harm when we allow it to go unregulated
and unchecked.
So, how should we respond? As we have noted, the Biden-
Harris EO initiated a whole-of-government approach and private
sector involvement to establish the United States as a global
leader in ethical AI. Within the Federal Government, the EO
directed more than 50 Federal agencies to take more than 100
different actions to guarantee responsible Federal use in the
over 700 use cases already implemented across 24 Federal
agencies. Every agency should now have a designated chief AI
officer and internal AI governance board, which should work
cohesively and collaboratively to manage the risks of AI while
prudently removing barriers to innovation.
In addition to the EO, President Biden established a
Blueprint for an AI Bill of Rights to guide equitable uses of
AI across the public and private sectors. It was another key
step to ensure that every new technology comes with guarantees
of civil rights, civil liberties, privacy, and equal
opportunity. The blueprint also included private industry input
from companies like Amazon, Anthropic, Google, Inflection,
Microsoft, and OpenAI, each of whom committed to strengthening
safety, security, and transparency as they proceed to innovate
new AI algorithms and explore potential use cases for the
technology.
As the Federal Government continues to explore AI, we
should look to partner with private sector partners to foster
continued innovation and adoption of a secure and trustworthy
AI. I plan to introduce a bill that responsibly accelerates the
use of AI by our civil servants who are entrusted with carrying
out our public-facing agencies' missions. AI must empower
workers, not replace them. We can achieve this future if we
build a robust educational foundation and one that benefits
both government and the private sector. Congress must invest in
programs that educate and train the next generation of skilled
AI workers, which the Chair has just mentioned, and they need
to thrive in the tech economy. Both the private and public
sectors will need digital native workers who are steeped in the
practices needed to put appropriate guardrails in place to help
AI achieve its goals.
Some of my colleagues and witnesses here today will say
that regulation limits innovation and stifles growth. It can,
but let me remind you of what our previous witness, Dr. Rumman
Chowdhury, a Responsible AI fellow at Harvard, said, ``Brakes
help you drive faster.'' This country has a strong history of
creating rules of the road for innovative industries in ways
that catalyze growth and foster trust in consumers. We look
forward to exploring how best we can achieve that balance, and
I look forward to today's hearing. Thank you.
Ms. Mace. Thank you, Mr. Connolly. I am pleased to
introduce our witnesses for today's hearing. Our first witness
is Mr. Adam Thierer, Senior Fellow for the technology and
innovation team at the R Street Institute. Our second witness
is Ms. Jennifer Huddleston, Technology Policy Research Fellow
at the Cato Institute. Our third witness is Mr. Neil Chilson,
head of the AI policy at the Abundance Institute, and our
fourth witness today is Dr. Nicol Turner Lee, Senior Fellow for
Governance Studies and Director of the Center for Technology
Innovation at Brookings Institution. Welcome, everyone. We are
pleased to have you this morning.
Pursuant to Committee Rule 9(g), the witnesses will please
stand and raise their right hands. This is where it gets real.
Do you solemnly swear or affirm that the testimony that you
are about to give is the truth, the whole truth, and nothing
but the truth, so help you God?
[A chorus of ayes.]
Ms. Mace. Let the record show the witnesses all answered in
the affirmative. We appreciate all of you being here today and
look forward to your testimony.
I would like to remind the witnesses we have read your
written statements, and they will appear in full in the hearing
record. Please limit your oral arguments this morning to 5
minutes. As a reminder, please press the button on the
microphone in front of you so that it is on, and the Members
can hear you. When you begin to speak, the light in front of
you will turn green. After 4 minutes, the light will turn
yellow. When the red light comes up, your 5 minutes has
expired, and we would ask that you please wrap up.
So, I will now recognize Mr. Thierer to please begin your
opening statement.
STATEMENT OF ADAM THIERER
RESIDENT SENIOR FELLOW, TECHNOLOGY AND INNOVATION
R STREET INSTITUTE
Mr. Thierer. Thank you. Chairwoman Mace, Ranking Member
Connolly, and Members of the Subcommittee, thank you for
holding this hearing and for the invitation to appear before
you. My name is Adam Thierer, and I am a Senior Fellow at the R
Street Institute where I focus on emerging technology issues. I
also recently served as Commissioner for the U.S. Chamber of
Commerce Commission on Artificial Intelligence,
Competitiveness, Inclusion, and Innovation.
My message here today boils down to three main points.
First, it is important to recall the foundational principles
behind the bipartisan National Framework for Digital Commerce
that Congress and the Clinton Administration crafted a quarter
century ago. Freedom to innovate was made America's policy
default. Lawmakers rejected the inefficient, old regulatory
models of the analog era, which constrained entrepreneurialism
and competition. We allowed new digital technologies to be born
free and to flourish without excessive micromanagement, and
then we used ongoing multistakeholder efforts and flexible
regulatory responses to address concerns. Europe took the
opposite path, and today, heavy-handed technocratic mandates
have ``regulated its way to last place,'' as a recent Wall
Street Journal headline observed. In fact, 18 of the 25 largest
digital technology companies in the world today are U.S. based,
while it is difficult to name any that are headquartered in
Europe. While some people have concerns about large technology
companies today, we should agree that it is better that these
firms are primarily based here in the United States instead of
China, Europe, or other countries or continents.
Further, there is a second point about the connection
between AI policy and broader national objectives. A strong
digital technology base is an important source of strength and
prosperity, so it is essential that our Nation not shoot itself
in the foot as the next great technological race gets underway
with China and the rest of the world. Consider this scenario.
When OpenAI launched ChatGPT in late 2022, it quickly became
the most rapidly adopted digital technology in history, and
competing U.S. services from U.S. developers followed quickly.
Had a Chinese operator launched a major generative AI model
first, it would have been a Sputnik moment for America.
Luckily, it is instead foreign nations who are today left
scratching their heads, wondering how America once again raced
ahead of them on digital technology. Wise policy choices not
only strengthen our economy and provide better products and
jobs, but also bolster our national security and allow our
values to shape information technology platforms and markets
globally. We need a national AI policy that is flexible and
pro-innovation to make sure our firms, workers, and values
continue to lead the world in this fashion.
This brings me to the Biden Administration's October
executive order. This wide-ranging, 100-plus page directive has
been praised by some as a logical response to congressional
inaction on AI, but many others have rightly noted that it
stretches executive authority over emerging technology well
beyond statutory limits and raises the danger of
overregulation. For example, the order flips the Defense
Production Act on its head and converts a 1950's law, meant to
encourage production, into an expansive regulatory edict
intended to curtail some forms of algorithmic innovation.
Twenty state attorneys general recently filed a letter with the
Department of Commerce noting how the order is ``about
regulating technological development, not about encouraging the
production of anything,'' and also objecting to its effort to
``centralize government control over an emerging technology
being developed by the private sector.''
The order also contains open-ended language about taking
steps to combat algorithmic discrimination and pushes the
Federal Trade Commission to get more aggressive in policing the
AI marketplace. These steps open the door to a new regulatory
regime for AI without any express authority from Congress.
While other provisions of the order are more reasonable,
Congress still needs to reassert itself to ensure that
administrative overreach is curtailed and that agencies adhere
to the Constitution and their congressionally delegated powers.
Instead of these arbitrary, excessive mandates, Congress
needs to craft an AI policy vision that does four things:
first, it is rooted in a flexible, risk-based framework that
relies more on ongoing, multistakeholder negotiations and
evolutionary standards that are more closely matched to rapidly
changing algorithmic technologies; second, which builds on
existing government powers on a sectoral basis instead of
trying to develop an entirely new regulatory superstructure for
AI; third, which preempts state and local government AI laws
that create confusing patchworks of conflicting mandates; and
fourth and most importantly, gives algorithmic entrepreneurs a
green light and avoids treating AI services as guilty until
proven innocent as the executive order does.
In sum, our Nation must create a positive innovation
culture and avoid trapping our AI innovators in a regulatory
cage if we hope to prosper economically and ensure a safer,
more secure technological base. It is essential that we strike
the right policy balance as we face serious competition from
China and other nations who are looking to counter America's
early lead in computational systems and data-driven
technologies.
Thank you for holding this hearing and thank you for the
consideration of my views. I look forward to any questions you
may have.
Ms. Mace. Thank you. I now recognize Ms. Huddleston to
begin her opening statement.
JENNIFER HUDDLESTON
TECHNOLOGY POLICY RESEARCH FELLOW
CATO INSTITUTE
Ms. Huddleston. Thank you, Chair Mace, Ranking Member
Connolly, and distinguished Members of the Committee on
Oversight and Accountability Subcommittee on Cybersecurity,
Information Technology, and Innovation. My name is Jennifer
Huddleston, and I am a Technology Policy Research Fellow at the
Cato Institute, where my research focuses primarily on the
intersection of law and technologies, including issues related
to the governance of emerging technologies, such as artificial
intelligence, better known as AI. Therefore, I welcome the
opportunity to testify today regarding the recent AI executive
order issued by the Biden Administration.
In this testimony, I seek to focus on two key points:
first, how the AI EO represents a significant shift in the U.S.
approach to AI policy and to technology policy in general; and
second, how the AI EO raises concerns about appropriate
separation of powers at a time when Congress is debating the
most sensible policy framework to consider for governing AI.
To begin with, the AI EO represents a significant shift
from a more permissionless approach to general purpose
technology and technology policy in general, to a more
permissioned or precautionary approach, such as those more
commonly found in Europe. While much of the conversation around
AI has been recently focused on the generative AI products like
ChatGPT or DALL-E that became popular with consumers in late
2022, AI and machine learning has been part of our lives for
far longer than many of us may realize. From tools that help
detect potential credit card fraud, to our talk-to-text or
autocompletes on our phones, to various things that help make
us find better and faster search results, we have all been
using artificial intelligence far longer than we may realize.
AI is helping fight wildfires and enabling stroke victims to
speak again, and it is estimated that AI could increase
productivity by 1.5 percent per year and global GDP by $7
trillion over the next decade.
All of this is to say, while much of the conversation is
focused on the potential harms of AI, we should not forget the
benefits as well. In fact, not all uses of AI can be predicted.
As we have seen with the internet, one of the things that
really allowed the U.S. to flourish and one of the reasons why
the light touch approach gave rise to so many of the wonderful
products that we have today, is that consumers and innovators
were able to decide what products were the best applications,
not government bureaucrats.
While the AI EO may be the most significant AI policy that
we have seen at an executive level, it is not the first
executive that has mentioned AI. In fact, both the Trump
Administration and the Biden Administration had comments on AI
and the importance that it may serve for economic growth and
its valuable tool in the future. Notably, the Biden
Administration's executive order looks less favorably on the
potential for a less regulatory approach to this technology,
and it suggests that there is a case for action amongst
agencies. It nudges agencies in a do-something direction more
so than prior administrations, particularly in its invocation
of the Defense Production Act.
This brings me to the second point today. The AI EO should
raise significant concerns about appropriate separation of
powers. Not only does this represent a change in the overall
approach to technology, it occurs at a time when Congress is
actively debating this issue and it occurs by executive order,
while we have seen many committees, including today's hearing,
on a wide range of topics in both the House and the Senate,
consider whether or not a further regulatory framework for AI
is necessary.
The most notable example of this is the AI EO's use of the
Defense Production Act to justify its provisions. This law was
originally designed to provide the executive with authority to
meet a national security cris, but the AI EO evokes the Defense
Production Act, not to respond to such a crisis, but, rather,
to require innovators of AI products, deemed high risk, notify
the government and submit to government-run red teaming
regarding the potential risk of their innovation. This
executive overreach cannot be presumed to have occurred because
there is a need for immediate action or because there is a lack
of attention on Congress' part, and the powers that it passes
on to the administrative state should be considered carefully
by Congress. As we know, once power is given to the
administrative state, it is unlikely to be returned.
The United States' light touch approach to the internet
helped enabled its global leadership and realize the economic
potential of this technology in the past years. This has
benefited both consumers and innovators and entrepreneurs. As
we encounter our next disruptive technology era with AI, we
must consider not only the risk, but also the benefits of such
technology, many of which we may be unable to predict. The
United States has a chance to distinguish itself from more
regulatory approaches once again and embrace an approach that
allows consumers and innovators to use technology to find
creative solutions to problems and needs.
I thank you for this opportunity to testify before you, and
I welcome your questions.
Ms. Mace. Thank you. I now recognize Mr. Chilson to begin
your opening statement.
STATEMENT OF NEIL CHILSON
HEAD OF AI POLICY
THE ABUNDANCE INSTITUTE
Mr. Chilson. Chair Mace, Ranking Member Connolly, and
Subcommittee Members, good morning. I am Neil Chilson, the head
of AI policy at the Abundance Institute. The Abundance
Institute is a mission-driven, nonprofit (c)(3) dedicated to
fostering widespread human prosperity by creating an
environment where emerging technologies, including AI, can
thrive. Thank you for having me here today to talk about the
executive order on artificial intelligence.
The artificial intelligence EO, as noted by the Chair, is
long. In fact, it is the longest regulatory EO in history.
According to data from the American Presidency Project, the AI
EO is 88 times longer than the median executive order. By the
numbers, it is the third longest executive order in American
history, but the two EOs that are longer are a 1951 executive
order containing the entire manual for the military court
martial procedures and a 1980 EO revising that manual. The AI
EO is also unusually regulatory because it directs actions by
dozens of agencies and Federal officials. It mandates 136
different deliverables, such as reports, guidance documents,
and dozens of new projects, processes, and plans.
In short, the AI executive order is unprecedented. Our
country's history includes many dramatic developments,
including civil war, mass industrialization, two world wars,
globalization, and global pandemics, yet no President has ever
issued such a long and detailed executive order to reorient the
Federal Government on a policy issue.
The executive order will generate sweeping activity across
the Federal Government, redirecting at least tens of millions
of tax dollars and hundreds of thousands of hours of government
effort. Some of that activity will be productive and
appropriate, particularly where it focuses on the government's
own use of AI, but the executive order overreaches in at least
two ways. First, the President lacks the authority to impose
the executive order's Section 4.2's obligations on private
companies. The executive order claims authority under the
Defense Production Act, which is a Korean War-era law intended
to reduce one very specific type of national security risk--
threats to ``the ability of the domestic industrial base to
supply materials and services that are needed for national
defense and disaster recovery.''
But the DPA does not authorize Section 4.2 for three
reasons. First, there is no threat to the ability of the
domestic industrial base to supply AI capabilities. The U.S.
leads the world in supplying AI capabilities. Indeed, the White
House, if anything, seems concerned with an oversupply of AI
capabilities. As the EO itself notes, AI capabilities are
advancing at rapid speed. Second, even if there were such a
threat, Section 4.2 will not increase production of AI
capability. Section 4.2 surveils an entire industry segment for
various potential risks unrelated to production. The only
production Section 4.2 will promote will be the production of
highly sensitive commercial and cybersecurity information from
companies to the government. Third, and most fundamentally, the
DPA simply cannot shortcut the constitutionally established
method of democratic lawmaking in the U.S. As we have heard,
Congress is actively considering 28 AI-related bills. The DPA
does not empower the President to skip ahead of Congress on
this.
The EO overreaches in a second way. The executive order's
definition of ``artificial intelligence'' is so broad that it
covers common everyday software, from social media content
moderation algorithms, to insurance models, to common consumer
and business financial tools. This broad definition means that
many of the regulatory actions spurred by the executive order
could apply not just to AI companies, but to any software
developer in industries, such as transportation, education,
healthcare, or energy. Even were it desirable to revise the
U.S.'s highly successful approach to software regulation, doing
so through a Presidential executive order is inappropriate.
Such a change deserves to be considered, refined, and decided
by Congress.
In conclusion, the executive order could have focused on
establishing a positive vision for a future of AI while
protecting civil rights from government misuse of these
technologies. Instead, it usurps Congress, abusing the DPA to
impose new regulations, and spurring regulatory action that
will affect the entire software industry. This overreach calls
for continued congressional oversight, democratic
accountability, and potentially legislative or judicial course
correction. Thank you, and I look forward to your questions.
Ms. Mace. Thank you. I now recognize Dr. Lee for her 5-
minute opening statement.
STATEMENT OF DR. NICOL TURNER LEE
SENIOR FELLOW, GOVERNANCE STUDIES, AND
DIRECTOR, CENTER FOR TECHNOLOGY INNOVATION
BROOKINGS INSTITUTION
Dr. Turner Lee. Thank you, Chairwoman Mace, Ranking Member
Connolly, and distinguished Members of the Subcommittee, for
the invitation to testify on President's Biden's executive
order on safe, secure, and trustworthy development and use of
AI. I am Dr. Nicol Turner Lee, Senior Fellow at the Brookings
Institution, which has a 100 year history of evidence-based,
nonpartisan research, and I thank you, Chairwoman, for the
mention.
With that being said in my brief remarks, I just want to
remind people that getting to the EO has been a long,
deliberate, participatory process, and it has been one in which
we have had several government actions proceeding and
surrounding it, such as the Blueprint for an AI Bill of Rights
released in October 2022, the National Institute of Standards
and Technologies AI Risk Management Framework released in 2023,
the securing of voluntary commitments by some of the top
companies in July 2023, and the OMB guidance released shortly
after October in November 2023. These actions reflect a whole-
of-governance governance approach, and they are really
important for us to achieve national guidance as AI becomes
both an asset and concern for our national security interests.
I want to also share that Congress must act quickly on many
of these proposals in our decision to maintain our status as
leaders in the global economy. Rather than say that this is
overreach, I consider these efforts to be preparation toward a
more responsible, inclusive AI ecosystem. This first formidable
action on AI under the Blueprint for an AI Bill of Rights
shared a nonbinding roadmap for the responsible use of
artificial intelligence. We then proceeded to have NIST, who
gave us the risk management framework as a multi-tool for
organizations to design and manage trustworthy and responsible
technologies that are meant to be voluntary, rights preserving,
nonspecific, use case agnostic. NIST is also going to release a
playbook that will be a companion to this, and we recently
launched the AI Safety Institute Consortium to bring
stakeholders together to jointly develop and diffuse best
practices, standards, and other things.
A few months later, as it was mentioned, the White House
secured voluntary commitments from some of the leading U.S.-
based AI companies that want to equally ensure safety,
security, trust with advanced systems. They are willing to not
only look at their own business models, but to find ways to
engage in public reporting of their system capabilities,
limitations, and guidelines for use. The most notable
advancement, so far, is the Robust Watermarking Solution that
these companies are working on together to ensure that we can
authenticate AI-generated content.
While the executive order may appear to be a variety of
issues packaged into one, it is intentionally designed to be
such. We have put in the work to make sure we have gotten to
this point, and the EO will not only pursue the eight
benchmarks that are outlined in its mandate, it also, on a
cumulative level, has pursued us to look at best practice for
AI use in criminal justice, education, healthcare, and other
thoughtful processes that develop an astute and ready
workforce. The actions to engage the Federal Government is
probably the most ambitious, yet necessary, action to confirm
our resilience among foreign actors and others who want to
leverage malicious attacks.
In my written statement, I opine more on the January 24
progress report issued by the White House, which suggests that
we actually are meeting many of those benchmarks, even in light
of some of the concerns of my colleagues. Going forward, if the
U.S. wants to be a leader in innovation, we must be responsibly
prepared to manage those risks. AI can be developed with
positive intentions, such as saving the climate, and
simultaneously lean into negative uses, such as a large-scale
generation and dissemination of misinformation and deepfakes,
activities that are quickly appending important democratic
institutions, like voting and elections infrastructure.
Moreover, the advanced capabilities of frontier models, like
generative AI, will only deepen these effects, particularly if
our government does not act quickly to get ahead of this
technology. Again, instead of seeing this as overreach, this is
a whole-of-government approach that is thoughtful,
participatory, coordinated, and have been percolating for a
matter of years.
More importantly, if Congress does not act, states will,
China will, and other nations who are not only leading us in
comprehensive legislation, but will soon be the standards that
are going to develop the AI ecosystem. That is why we have
urgent assignment to move and legislate on what is actually in
the EO. More so, instead of having a patchwork of state laws
and local provisions, it is important that Congress be
concerned about these interests of AI simply on the national
security interests, simply on the public consumer protection
concerns, and more so because we want to be the leaders in this
technology.
I urge Congress to consider that we already have many
proposals that have some bipartisan support, like the National
AI Commission Act, the Protect Elections from Deceptive AI Act,
and let us move so that we do not have to continue this
conversation again.
Ms. Mace. Thank you so much. I will now recognize myself
for 5 minutes.
Generally, my questions will be for the entire panel. I am
a big fan of ``yes'' or ``no,'' so if you want to elaborate, I
would just ask that you make it quick because I would like to
hear from a lot of voices here this morning. The executive
order is actually one of the longest EOs ever written at over
100 pages, and it absolutely will encumber tech in a lot of
ways, as we heard from our witnesses this morning.
My first question is, the Commerce Department could not
protect Secretary Raimondo's own email account from being
hacked last year, yet this EO requires firms to share with the
Agency on a daily basis the crown jewel secrets of the most
powerful AI systems on earth. First question: can we trust
Commerce to ensure this highly sensitive data does not fall
into the hands of China or another foreign adversary? Yes or
no.
Mr. Thierer. No.
Ms. Huddleston. I think we have seen that there is a need
for a greater discussion of improving cybersecurity, both in
the government and beyond.
Mr. Chilson. No, and we should not have to.
Dr. Turner Lee. Yes.
Ms. Mace. OK. Thank you. According to the executive order,
the over 100-page executive order, irresponsible use of AI
could exacerbate ``social harms,'' including,
``disinformation.'' Should we trust the government to be the
ultimate arbiter of what is disinformation and what may cause
social harm in AI systems? Mr. Thierer?
Mr. Thierer. No.
Ms. Huddleston. No, and we should be concerned about the
First Amendment approach of doing such.
Mr. Chilson. Absolutely not.
Dr. Turner Lee. I will say yes, and government, with our
civil society, industry partners, alongside of us.
Ms. Mace. Should AI model developers have to give the
government all their test results and test data, even those
concerning politics or religion? Mr. Thierer?
Mr. Thierer. No.
Ms. Huddleston. I think it is a highly concerning proposal
with significant consequences for innovation.
Mr. Chilson. No.
Dr. Turner Lee. You know what I am going to say.
Ms. Mace. You are going to say yes.
[Laughter.]
Dr. Turner Lee. Especially in high----
Ms. Mace. You are going to be the one dissenter this
morning.
Dr. Turner Lee. Yes, especially in high stakes.
Ms. Mace. What are some of the, just very quickly, the
risks if companies are giving their test data over to the
government? Mr. Thierer?
Mr. Thierer. There are security risks, of course. There are
also concerns about how there might be speech meddling of
various types, the sort of jawboning that could be associated
with that sort of heavy-handed approach.
Ms. Mace. What would the government do with such
information potentially?
Mr. Thierer. Well, it depends. We know in the past, there
has been efforts by government authorities to utilize such
information to try to curb certain types of behaviors or to try
to intimidate certain people to do things against their will
and without due process.
Ms. Mace. Ms. Huddleston?
Ms. Huddleston. Similarly, I think there are concerns about
how this could have an impact on speech as well as innovation
more generally, with the idea that innovators would have to
seek permission from the government before engaging in their
innovation, rather than having it play out in the marketplace
of ideas.
Mr. Chilson. I think one of the other big effects other
than the government misuse would be the chilling effects it
would have on people using tools like this to say things that
they think the government might not want them to say.
Dr. Turner Lee. And I would just suggest that I think we
are conflating surveillance technology and how we actually look
at government on the surveillance side versus what the EO is
actually dictating, which is accountability, and the AI
training data, and the test beds as we look at high-stakes
applications. So, I would like us just to clarify that.
Ms. Mace. OK. And then, well, my last point is, you know, I
think we can all learn something, probably from recent history,
the FBI's interpretation of the Hunter Biden laptop as Russian
disinformation, and we had a number of over 20 former intel
officers and folks that wrote a letter saying it was Russian
disinformation. Come to find out, it was not. I think that is a
concern that a lot of folks have on what the government will do
with data, what the government will do with information, what
the government will do with testing information, algorithms,
code and programming product, et cetera.
The EO requires companies even considering developing dual
use foundation AI models to report to the government on an
ongoing basis again about their most sensitive business
secrets. Could the justification for using the DPA here be used
in the future to demand highly sensitive plans and data from
firms in any emerging technology field? Mr. Thierer?
Mr. Thierer. Yes, it could, and we should avoid it for that
reason.
Ms. Mace. Ms. Huddelston?
Ms. Huddleston. Yes, and it is concerning with the power
that would give to the administrative state over technology,
more generally.
Mr. Chilson. The EO offers no limiting principle on the use
of the DPA, and so I think we could expect that people will
continue to walk down this road.
Dr. Turner Lee. And I would just suggest that the
invocation of the DPA is coming in absence of congressional
action. If Congress were to provide some parameters on how we
exercise some of the, you know, principles that are embedded in
the EO, as well as things like congressional activities that we
want to actually foster, I do not think we would have that
problem. Congress just needs to act and legislate.
Ms. Mace. All right. Thank you. I want to thank our
participants on the panel today. We appreciate your time, your
insight and expertise, and I will now yield 5 minutes to my
esteemed colleague from Virginia.
Mr. Connolly. Thank you. Mr. Chilson, I am a little
puzzled. You spent a lot of time criticizing the EO because it
was too long, and I guess sitting up here looking at a very
complex subject that has never been addressed before, and that
many other people sat where you are sitting, Republican
witnesses, chosen witnesses, all praise the EO as, yes, it
gives us a framework we can work with, and they were not
worried too much about overreach, you know. They felt it gave
us a platform we can build on as we learn more, as we
experience more. Why are you so bothered by the fact of its
length?
Mr. Chilson. So, I cannot speak for what other people might
think about the EO. I can only speak for my experience in
Federal Government and watching the regulatory process. What
concerns me about the length of the EO is its unprecedented
nature because it looks like legislation. If you slap the bill
number on top of it----
Mr. Connolly. OK. Can I interrupt you because I get that
point, too, and I made a note of that. You went on to say it
actually usurps the role of Congress. You said that. Well, you
know, we do have a government with three branches, and we are a
co-equal branch of government. So is the executive. And when
one of those branches fails to act, that creates a vacuum that
almost demands the others act, depending on the urgency of the
situation. Now, it is nice to say, and I am a big champion of
legislative prerogatives, and I believe Article I is Article I
for a reason, and Article II is Article II for a reason, namely
we are supposed to be the predominant arbiter of government.
That was certainly what was in Madison's mind, but that is a
different conversation.
So, in the first session of the 117th Congress, we signed
into law about a hundred bills, very far reaching in some
cases, very visionary. In this Congress, 31, and half of them
are post office namings. This is not a serious Congress. You
mentioned, I think, 28 bills addressing AI. Not one has become
law, and given our pace, it is unlikely any of them will become
law in this Congress. We are not doing anything, and when that
happens, it seems to me the President has an obligation to
address an urgent and imminent subject like AI. And so, even
though I share your passion about the prerogatives and
responsibilities of the legislative branch, in this particular
case, I have to defend the executive branch. They have not
usurped Congress. They have actually done what Congress ought
to do but is not going to do because we are not doing our jobs
up here. In fact, we are not only not doing our jobs, we are
actually regressing.
So, in the appropriations bill, Nancy and I and others are
going to vote on this week, apparently, they have zeroed out a
congressionally created fund--by the way, I worked with Will
Hurd on this, a Republican--the technology management fund,
because apparently we do not need any more investment in
technology. You know, we do not need more cyber capability. We
do not need more AI training. We do not need any of that stuff.
We do not need to protect data bases that you are worried about
being compromised. And so, we are going to zero out the
technology management fund created by Congress. So, we are
going backward, we are not going forwards, and I just
respectfully disagree with you. I do not think the executive
order, I do not care how long it is myself, and I do not think
that is a particularly viable critique. And with respect to
usurping Congress, I do not think so.
Now, the third critique I think you had in your testimony
was overreach and echoed by your two colleagues on the other
side of the panel. I want to give Dr. Turner Lee an opportunity
to address that one. So, is it overreach? I mean, should we be
worried that they have gone too far, and they are intruding in
our lives, and they are going to compromise the ability of AI
and all of its promise?
Dr. Turner Lee. I would disagree with it being overreach.
When I started in this space back in the early 2000's--I have
been in technology since then when we were looking at regular
data bases and stuff--we did not have any framework for
predictive decisionmaking, and we had many conversations on
algorithmic bias. We were seeing that people were having equal
opportunities foreclosed simply because an algorithm made a
decision. That was the very first basic step at looking at
algorithmic discrimination and stuff like that. Now we have
advanced capabilities through frontier models that are actually
extracting data-text, voice, images--in ways that actually are
so opaque and less transparent that we need more guidance. The
technology continues to outpace policy in that matter. With
that being the case, to your point, I think we have had many
congressional bills that have come to the Floor, but they have
been too late, so we have had to relitigate and remitigate what
those bills are. In addition to that, we have been slow to the
pace when it comes to data privacy protections, reevaluating
our civil rights framework, things like that. I, in no way,
think this is overreach.
As a scholar at Brookings who is interested in trying to
figure out proactive, evidence-based strategies to move
forward, this is actually preparation. Without such
preparation, as I said in my testimony orally as well as
written, other countries are going to define the landscape for
AI regulation, and we will be subjected to their rules, not
just on the behavioral aspects, but also on those sides of the
technical cadence.
Mr. Connolly. Thank you. Madam Chair, I yield back.
Ms. Mace. Thank you. I will now recognize Mr. Timmons for 5
minutes.
Mr. Timmons. Thank you, Madam Chair. I guess, first, I just
want to respond to my colleague's, across the aisle, indictment
of this Congress. Us not doing anything this Congress is a
response to my colleagues across the aisle spending about $7
trillion last Congress, and it has really caused my
constituents a lot of problems. Inflation is through the roof.
The cost of energy, groceries, interest rates are up. It is
really costing Americans a lot. And so, we are struggling with
our $34 trillion in debt. We are struggling with the fact that
every hundred days we add a trillion dollars to our national
debt, and really, our fiscal situation is out of control. So,
we are very concerned about that, and we are going to try to
find a path forward that is sustainable for the American
people, that will give our kids and grandkids the American
Dream for generations to come, but I am very fearful that that
is not going to work out. So, we are going to focus on that.
That said, we do need to address AI, and just because
Congress is unable to address AI as quickly as we should does
not give the President the right to legislate for us. It is
just a really bad idea. You know, this has been tried again and
again. I guess first, Mr. Chilson, who wrote this? I mean, this
is very technical. It is very long as we keep saying, and, I
mean, who wrote this?
Mr. Chilson. I mean, it went through a White House process.
I do not know many details about that process. It looks like a
bunch of people wrote it.
Mr. Timmons. It is extremely technical, right? I mean, it
is extremely technical, so, I mean, people that have expertise
that probably have an interest in a regulatory structure being
placed on this. I mean, do you have any idea what groups were
involved? Is there any transparency to that?
Mr. Chilson. I do not. I think that is a good question for
Congress to ask and an oversight committee to ask, and I do not
have good information on that. I know there were a lot of
participants. There were probably a lot of people who are
asking for different things to be included in this, and that
is, in part, why it ended up so long.
Mr. Timmons. So, all the reporting requirements, does that
not chill innovation? I mean, there are a lot of proprietary
approaches to this. Is it going to chill innovation?
Mr. Chilson. I think it could chill innovation in a couple
of different ways. One, companies, when they are thinking about
what sensitive data that they are going to have to report to
the government, they are going to have to make a tradeoff.
Like, are we going to grow big enough to meet the caps that put
us over this reporting threshold, or are we going to stay under
that and limit ourselves artificially in order to not have to
comply with these specific rules? And I think that would be to
the detriment of U.S. leadership in AI.
Mr. Timmons. A lot of these companies are global. I mean,
could they not just move their development overseas outside the
jurisdiction of the U.S., or does the Defense Production Act
somehow extend beyond our borders?
Mr. Chilson. They absolutely could move overseas, and I
know that there are jurisdictions that are actively recruiting
AI startups and AI companies to move to their jurisdictions,
promising them less constraining regulatory environments.
Mr. Timmons. Dr. Lee, you used the term ``nonbinding.'' You
were not referencing this executive order, were you?
Dr. Turner Lee. In terms of the nonbinding reference, we
know that much of the content that was in the AI Bill of Rights
was nonbinding. It was more voluntary. And up until the
executive order, as you all are aware, we really do not have
any binding requirements unless the DPA is used for the test
bed.
Mr. Timmons. But----
Dr. Turner Lee. It only applies to certain aspects of the
EO, as we know.
Mr. Timmons. OK. So, my understanding is that there are
criminal penalties for not abiding by the DPA, and this is
using those authorities, so would it not be a criminal offense
that would result in jail and a fine should a company not
comply? And I guess another question is, who would that even
apply to? Would that be, like, the CEO? Would it be the board
members of the company? How does that work?
Dr. Turner Lee. So, I am with you. There actually has to be
more clarification on how that enforcement strategy looks like,
but remember, the regime of the DPA applies to only specific
aspects of the EO itself. And to your question, if I may,
respectfully, who wrote this was the American people, right? It
was accumulation of all these activities leading up to the
actual EO itself. So, I wanted to clarify because I kind of
have some insight into some of the stakeholders that
participated.
Mr. Timmons. I mean, I think some of the technical
expertise of this--look, I got a master's degree in
cybersecurity, and I understand this as well as anybody.
Dr. Turner Lee. Yes.
Mr. Timmons. And I am reading some of this, and, you know,
it requires a quantity of computing power greater than 10 to
the 26 integer. I do not think the American people have any
idea what that means, so with all due respect, the American
people did not write this. Mr. Thierer, do you have any
understanding of how this would be enforced in regard to
noncompliance? I mean, would the CEO or would board members go
to jail? Like, how would that work?
Mr. Thierer. I think that is a great question. Recall that
in the letter that the 20 AGs sent, they actually referred to
this ``opaque and undemocratic process'' by forcing AI
developers to submit information, but it was unclear to the AGs
themselves, and they asked the Department of Commerce, like,
what is going on here? So, we do not have answers to your
questions, Congressman.
Mr. Timmons. So basically, we took a problem and made it
worse. It seems that way.
Mr. Thierer. I think so.
Mr. Timmons. But again, Congress does need to act, to be
fair, so maybe we should get our act together and address this
in a way that can facilitate innovation and keep the United
States on the forefront of being the best economy in the world.
Thank you, Madam Chair. I yield back.
Ms. Mace. Thank you. I will now turn it over for 5 minutes
to Ms. Pressley.
Ms. Pressley. Thank you, Madam Chair, and thank you to our
witnesses for being here today. As a Member of the Financial
Services Committee's bipartisan Working Group on Artificial
Intelligence, I have no doubt that while AI presents
opportunities for progress, it also poses significant risks,
from undermining our privacy, to inciting political violence,
to spreading disinformation. Congress has been slow to act,
forcing the Biden-Harris Administration to take executive
action to enforce standards and guardrails. The AI EO does just
that, and to suggest that the White House is overstepping,
especially when just last week, this Subcommittee heard
devastating testimony on AI's infringement on the privacy and
civil rights of women and girls. So that overreach
characterization is absurd, in my opinion.
Dr. Turner Lee, in what ways can AI pose disproportionate
threats to people from marginalized backgrounds?
Dr. Turner Lee. That is an area that I spend a lot of time
with, and I think the effects on marginalized populations are a
couple of things. One, the lack of transparency of AI systems,
and particularly how they factor into predictive decisionmaking
or eligibility concerns, can foreclose on equal opportunities.
People do not know what those factors are that are going into
credit decisions, housing decisions, criminal justice
decisions, and the like. I would also say that people of color
are disproportionately impacted by deepfakes and
misinformation. The lack of transparency, actually, which is an
issue. Deepfakes affect anybody in any state and any party when
you actually look at it, but the lack of transparency
particularly affects communities of color who have less agency.
And then finally, I would just say criminal justice. I just
spent a year and a half with the National Academies on facial
recognition use in law enforcement, and in that application of
AI, we also see a lot of vulnerabilities as well.
Ms. Pressley. Thank you. Yes, AI algorithms trained on
skewed, inaccurate, or unrepresentative data magnify human
biases, lead to discriminatory outcomes. The previous
Administration, for example, has an abysmal record of using
technology to incarcerate and to persecute communities of
color. The Trump Administration used AI to identify legal
protesters during the George Floyd protest, to employ racist
algorithms with Immigration and Customs Enforcement to profile
Muslims entering the country, and haphazardly arrest Chinese
Americans during its China initiative. Meanwhile, President
Biden's executive order takes unprecedented action to allow
innovation while protecting people's privacy and civil rights.
Dr. Turner Lee, are the steps outlined in the Biden-Harris
Administration's EO sufficient to address biases in AI that can
lead to discriminatory outcomes?
Dr. Turner Lee. I wholeheartedly agree. I applaud this
Administration for including words like ``equity'' and
``parity'' as part of the EO in very outright ways so that we
address this issue front hand. I also think, to your point and
to the earlier conversation from my colleagues around the
government use of AI, it is very clear in the EO this
distinction between government surveillance that is used for
malicious intent by government, versus resiliency, which is the
Federal agencies just having clearer pathways on their use of
AI generally, whether it is in benefits decisions, criminal
justice decisions and actions, and so forth.
Ms. Pressley. Thank you. And, Dr. Turner Lee, what elements
of the EO can Congress strengthen to ensure that advances in AI
technology are not used to further involve people with the
criminal legal system?
Dr. Turner Lee. I think that Congress can take some steps,
and there has been some bipartisan support around the use of
facial recognition technology and how we actually not
necessarily ban it, but we have some guardrails that make sense
for various communities. I think Congress can also act on data
privacy legislation. That legislation will allow some sense of
guidance on what data can be collected, and in the area of
biometric collection that can also safeguard communities of
color. I think conversations on election and AI infrastructure
and architecture should be of concern, and it has been on a
bipartisan level. I think all of us are concerned about the
integrity of our elections based on artificial intelligence and
generative AI. So, I think there is a host of them. I am happy
to share more of those with you, Congresswoman, going forward.
Ms. Pressley. Thank you, and, you know, certainly we have a
responsibility to be innovative in our efforts in order to
build reliable protections for everyone, especially those who
have historically been left behind or targeted. So, I invite
all my colleagues to link arms and minds, if you will, in
carrying out that work. Whether it is the use of facial
recognition technology to criminalize people of color, deep
fake pornography to degrade women, or biased algorithms to keep
vulnerable community members from accessing critical resources,
existing equity concerns are at risk of being worsened for
people in my district, the Massachusetts 7th, and across our
country. Thank you, and I yield back.
Ms. Mace. Thank you. I will now recognize Mr. Burchett for
5 minutes. I am looking forward to your questions on AI, sir.
Mr. Burchett. Thank you, Chairlady. Mr. Thierer and Mr.
Chilson, what is the historical background of the Defense
Production Act?
Mr. Thierer. Well, it was put in place, sir, to make sure
that America had the proper productive capacity of the
environment.
Mr. Burchett. I realize that, but, I mean, I want to know
the background. What caused it to be in place?
Mr. Thierer. A concern about the lack of a productive
capacity in certain sectors that the Federal Government felt
were necessary to achieve various national security purposes.
This was, of course, in the 1950s, a different time.
Mr. Burchett. Right. That is what I was getting at, the
Korean War.
Mr. Thierer. Yes, a long time ago.
Mr. Burchett. Yes, sir. And the primary purpose of the
Defense Production Act is to allow the President to direct the
production of materials and goods. Is that correct?
Mr. Thierer. Yes, that is correct. Yes, that is correct.
Mr. Burchett. OK. What materials or goods does Executive
Order 14110 direct companies to produce?
Mr. Chilson. Documents containing highly sensitive
commercial and cybersecurity information.
Mr. Burchett. OK. And what national security concerns exist
regarding AI that justifies using the Defense Production Act?
Mr. Chilson. Well, I think there are national security
concerns around AI. We have heard a lot of talk about, you
know, the rivalry with China and the importance of staying
ahead, so there are concerns there. But as for ones that
directly address the kinds of threats to interrupted production
that the Defense Production Act is looking for, again, as Adam
said, it turned the DPA on its head, which the DPA is to allow
the government to spur production. And the executive order uses
the DPA in order to discourage production on some levels, in
part by imposing additional regulatory burdens on people who
are producing at the highest level.
Mr. Burchett. Has the Defense Production Act been used to
extract information from companies rather than to encourage
production? Either one of you all.
Mr. Thierer. Not that I am aware of.
Mr. Chilson. I have heard that there has been already an
immediate request based on this use of the Defense Production
Act, but I think in the past, I am not aware of another one.
Mr. Burchett. Do you think using the Defense Production Act
to regulate artificial intelligence is a bit of an overreach,
and would Congress be better suited to regulate artificial
intelligence?
Mr. Thierer. Yes, Congressman, I think that is right. The
authority begins here to decide what the Defense Production Act
should do, and I think now we are witnessing pretty excessive
overreach of the statute.
Mr. Burchett. Do you all think that this executive order
could stifle artificial intelligence innovation?
Mr. Chilson. I do, and I think the use of the DPA here
undermines some of the other important goals that Congresswoman
Pressley was pointing out about government uses and the risks
of government use of AI.
Mr. Burchett. Ma'am, you were shaking your head.
Dr. Turner Lee. May I respond? Yes. I actually disagree
with that. I think, and going to my colleague here, who I have
known for many years, I think what the Congresswoman was
talking about does not require the use of the DPA, in all
honesty. It actually just requires transparency, disclosure,
that kind of stuff. I think the DPA was actually exercised
based on just giving some push to us to do something as a
national economy so that we make sure we are not behind others,
particularly China, when it comes to AI.
Mr. Burchett. Ms. Huddleston, you have not responded. Would
you like to?
Ms. Huddleston. I would agree with Mr. Chilson that I do
think there are significant concerns about how the executive
order could stifle innovation at a time when AI is still just
emerging and we are just starting to understand the potential
beneficial applications of it, as well as the potential risk.
Mr. Burchett. Do you all think that the executive order
strengthens the U.S.'s ability to maintain its lead over China?
Ma'am?
Dr. Turner Lee. I do, and, again, responding to my
colleagues, when we talk about stifling innovation and
invention in this country when it comes to AI, I think we have
two different conversations going on. One is a conversation
around the efficiency and use of AI in areas like, you know,
occupational careers, different sub-stacks, technological
applications. The other was around the sociotechnical
application. How does the public interest benefit from the use
of AI?
Mr. Burchett. Right.
Dr. Turner Lee. And I would just urge us to sort of not
conflate those two areas.
Mr. Burchett. Well, how does it do that?
Dr. Turner Lee. When we have an informed populace that
understands that technology is embedded in basically everything
that we are doing today, our informed populace can make
decisions that actually benefit their quality of life. When
they do not know that these technologies or AI-generated
content is happening, we are actually stifling our ability to
move this Nation into a space where, to your point, we can stay
competitive with our rivals.
Mr. Burchett. Yes, my biggest fear with this, again, is we
do not understand it. Heck, I do not understand it, and
Chairlady Mace probably understands it, and my colleague across
the aisle probably understands it. But here again, we are going
to start regulating something we do not understand because we
are government and we are supposed to, and then, again, it is
just like cryptocurrency and everything else, dadgummit. We
will end up hurting it, you know, so that is my concern.
Chairlady, I yield back none of my time. Matter of fact, that
is a negative amount of time, so I do not know if that
penalizes me on the next or not.
Ms. Mace. You did good, Mr. Burchett. I will now yield to
Mr. Burlison for 5 minutes.
Mr. Burlison. Thank you. Dr. Lee, would you say that
democracy is important in the United States?
Dr. Turner Lee. Yes.
Mr. Burlison. Would you say at times it is a threat?
Dr. Turner Lee. Yes.
Mr. Burlison. I would agree. Let me ask you this. Would you
think that an authoritarian state that does not represent the
elected people is a threat to democracy?
Dr. Turner Lee. I am happy I do not live in an
authoritarian state, in a democracy that we have here in the
United States, so I cannot tell you from personal experience.
Mr. Burlison. Well, so, I humbly disagree because
unfortunately, what we are seeing here, in my opinion, is an
authoritarian move. This new executive order is not being
conducted by the legislative body, people that were elected to
represent the people of the United States. It is being written
by people who have never run for office, to my knowledge,
probably never run for office, do not have to answer to any
voters whatsoever. So, Mr., is it Thierer?
Mr. Thierer. Yes.
Mr. Burlison. OK. In an analysis of the executive order
last year, you stated that, ``The unilateral and heavy-handed
administrative meddling in AI markets could undermine America's
global competitiveness and even the Nation's geopolitical
security, if taken too far.'' Has this executive order gone too
far?
Mr. Thierer. It very well could. You know, Congressman,
just this week, Saudi Arabia announced historic investment in
its AI capacity, something like $40 billion. Last September,
the Government of the UAE came out with an open-source AI model
that is 2.5 times larger than America's largest open-source AI
model. So, it is not just China we face off against, it is all
sorts of countries. Russia just developed one of its biggest
supercomputers. If this executive order shoots ourselves in the
foot as a Nation and holds back our innovative capacity, that
has massive ramifications for our competitiveness and our
geopolitical security.
Mr. Burlison. Yes. My other question has to do with the
fact that the executive order establishes an HHS AI task force
tasked with developing a strategic plan to regulate aspects of
AI in the healthcare industry, including research and
discovery, drug and device safety, healthcare delivery and
financing, and public health. Could this lead to an onslaught
of additional regulations?
Mr. Thierer. Absolutely, and we were already seeing it. We
should keep in mind our Federal Government is massive; 438
Federal departments, 2.2 million civilian workers working at
them. Every one of these agencies is interested in taking a
look at AI. This executive order essentially gives them the
green light to do so and says go for it without any express
congressional intent.
Mr. Burlison. As someone who worked in the healthcare IT
industry for 20 years, I can tell you that this place does not
aid, was not helpful in improving the lives of the American
people when it passed, under the American Recovery and
Reinvestment Act, the meaningful use criteria that every
software electronic medical record system had to accommodate in
order to continue to receive full reimbursement from Medicare.
The outcome, my statement, the outcome is basically proven
that, and that is that what that regulation did was shut down
many electronic medical record companies across the United
States, which forced doctors to consolidate, change their
records, migrate them to a new platform, or to stand up a
platform altogether. They were happy with a paper chart. Would
you agree with me that that outcome, creating what is basically
a duopoly in the healthcare IT space, is not good for doctors,
not good for patients, not good for consumers?
Mr. Thierer. Yes. Well, of course not, and, of course, this
effort by the Administration is just going to add more
compliance costs and regulations on top of it. I know you
mentioned this at the last hearing on this, Congressman, that
these sorts of burdens can compile and buildup on small
innovators and force them to move or get out of the field.
Mr. Burlison. Thank you. And, you know, at the end of the
day, it was said that if we do not take action, the states will
take action. Well, I seem to recall that the United States
Constitution and this system of government did not create the
states. In fact, it is the other way around. The states created
the Federal Government. And in the Tenth Amendment, it
specifically says, ``The powers not delegated to the United
States by the Constitution nor prohibited to it to the states
are reserved to the states, respectively, or to the people.''
So, my question to you, ``left to the states,'' wouldn't it be
better to have a microcosm of experiments, especially in a
field that we know so little about at this point in time?
Mr. Thierer. It depends on the rules. There are a lot of
proposals out there, a huge increase in the sort of compliance
burden if we have too much of a patchwork. The leading law on
AI hiring right now in the Nation is from New York City. Not
New York State. New York City. You can imagine if every city
has its own plan on AI, that could be a problem.
Mr. Burlison. Thank you. My time has expired.
Ms. Mace. All right. We are waiting on one more Member who
is coming, so I am going to ask a few questions that I did not
get to ask earlier, so I will now recognize myself for 5
minutes.
So, the executive order came out at the end of October, and
it said that Commerce would have to implement and get moving
after 90 days, which would have been the end of January. Does
anyone on the panel know if Commerce is doing anything related
to this executive order at this juncture today, middle to end
of March, soon to be April? Does anybody know?
Mr. Chilson. They certainly are doing a lot, actually.
Ms. Mace. Mm-hmm.
Mr. Chilson. There are several rulemakings and other
proceedings that are ongoing. NTIA, which is part of the
Department of Commerce, has a rulemaking on open-source models,
and so there is a lot of swirl at NTIA, and I think across
Commerce, so yes.
Dr. Turner Lee. And I would echo that, Chairwoman, that
Commerce has actually started, like other Federal agencies, to
go deeper into these issues. So, there are activities as well
as comments that will come out shortly, I am sure, on that.
Ms. Mace. Any movement on the DPA provisions of the EO?
Mr. Chilson. I have only heard that there was a request to
some companies in mid-December that they start filing
information. The understanding of the DPA provisions as the
executive order came out was that none of the current models
met the threshold for reporting yet, and so that request that I
have heard has gone out, it is not quite clear why that request
had gone out, and so----
Ms. Mace. Because they had not met the threshold.
Mr. Chilson. There is language in the DPA provision----
Ms. Mace. Mm-hmm.
Mr. Chilson [continuing]. That says if you are even
thinking of getting over the threshold, you also have to
report, and maybe that is----
Ms. Mace. And how do they report? How are companies----
Mr. Chilson. I do not know what the exact procedure is. The
executive order does not lay that out in specific detail. I
think it would be up to whatever the requests are that are
coming in. In fact, the executive order does not give a ton of
detail about what specific information. It gives general
categories, but I am hearing that they are asking for quite a
lot, including a lot of sensitive information.
Ms. Mace. It gives them, basically, broad authority to do
whatever. Mr. Chilson, it gives them broad authority, then
because of that vagueness?
Mr. Chilson. I think so.
Ms. Mace. Mm-hmm.
Mr. Chilson. Broad authority to request the information
that fits into those buckets that are in the executive order.
Ms. Mace. Gotcha. I think Dr. Lee wanted to chime in.
Dr. Turner Lee. And just respectfully, we are also finding,
though, with the progress report as well as my own research,
that various Federal agencies, including NTIA, are becoming
more transparent with their processes. So, I assume that we are
actually going to see much more openness around what is
happening around the executive order as they have been charged
to take on certain aspects of that, Chairwoman.
Ms. Mace. OK. And then this is a question for the entire
panel. How does this EO stifle innovation, limit potentially
investments in AI in the United States?
Mr. Thierer. Well, I will just go back to the AG letter
that really nailed it because they asked the Department of
Commerce to answer some questions like you have just asked,
Congresswoman. And when they referred to that opaque and
undemocratic process of forcing AI developers to submit
information for review behind closed doors, they also then
talked about the danger of a bureaucratic and nebulous
supervisory process that will discourage development, further
entrench large incumbents, and do little to protect citizens.
They asked a whole series of questions like you are asking to
the Administration. I have not seen any answers back.
Ms. Mace. Ms. Huddleston, how will this stifle innovation,
stifle investment, because I believe that it will, having these
encumbrances and burdens.
Ms. Huddleston. And even in the places where the EO is
vague, it signals to innovators that the government is
expecting them to take a seek permission first approach,
particularly for large, significant models and as this
technology evolves. It signals that the Administration is kind
of presuming this technology is a risk until proven otherwise.
We saw during the internet era how this more precautionary
approach for a general-purpose technology played out in Europe,
that it led to not seeing the kind of development of companies
that we saw in the U.S. And while there are many other factors
to that, such an approach that requires innovators to seek
government permission rather than consumers and innovators to
decide in the market how a technology can progress may stifle
innovators from going into certain areas, or may give rise to
investors' concerns about whether a technology will be allowed
to fully develop, particularly for those dual-use technologies.
Ms. Mace. Mr. Chilson?
Mr. Chilson. So, in addition to the DPA chilling effects
that will happen under the authorities there to some of the
largest models, I really do think, and to the earlier
Congressman's point, that the very vague definition of
``artificial intelligence,'' which to my reading and I think
some other experts' readings, could sweep in things like
formulas in complicated spreadsheets. This is not ChatGPT-style
AI. We are talking about software that people use every day and
that industries use every day. The Federal Government has now
been told, including HHS, hey, go make sure that all of these
AI pieces of software are working well, that they are meeting
the appropriate levels of quality. That has the potential to
unleash a lot of regulation, not just on what we think of as
AI, but software development generally, which would be really a
sea change in how the U.S. approaches software development.
Ms. Mace. I know Dr. Lee is going to disagree here. So why
does it not, in your opinion, because to me it is so obvious. I
am just curious on your perspective, why it would not stifle
innovation or not stifle investment.
Dr. Turner Lee. Well, as one who has been in the technology
space for more than 30 years, I clearly know when we see these
technology disruptors stifle innovation. I think in this case,
what we are going to see is improved certainty and baselines
for companies to better participate in the AI economy. Right
now, it will provide some behavioral guidance as well as some
product design guidance that I think will be helpful as we
actually, again, leverage AI as a national security interest.
The other thing I want to just continue to refresh and remind
us, that this is just not about the DPA, right? It is about
creating an AI workforce. It is about making sure that we have
the right research. It is about ensuring that our Federal
agencies are resilient, and it is about making sure consumers
are protected at any time, at any point in which they are
engaging these technologies. So, with that, I think the
certainty will definitely not stifle innovation in many
respects and will actually help us to innovate better.
Ms. Mace. With regards to DPA, do we see this applied to
any other technology? Is that a thing?
Mr. Chilson. The DPA has been used to ensure the supply of
materials that are needed for defense and for disaster recovery
on a wide range of technologies. We have never seen it applied
to AI, and we have never seen it applied not to ensure that
there is sufficient production for defense, but more for a
regulatory purpose, like it is being used here, to make sure it
is safer or that it is more limited than it is. So, this is a
very unprecedented use of the DPA.
Ms. Mace. And then, you know, I sit on not just the
Oversight Committee and Chair of the Cybersecurity Committee
here on the Oversight Committee, but I sit on House Armed
Services. I am privy to a lot of briefings, classified
briefings, about what our adversaries are doing, what they are
up to, technology wise, even AI. And, you know, one of my
greatest concerns from a national security perspective is
advances that some of our adversaries are. Like, you know, I
feel China is right on our heels, and I do not want to, I
guess, limit our ability to keep up with the technology.
And, you know, I have met with a lot of different tech
firms and seen a lot of the benefits of AI. One of them, for
example, is a company that maps the world every single day with
200-plus satellites in space. They can map every inch of the
earth every single day. Now, a year ago, this company, they
were doing great. If you need to find a little widget or target
or something on the map on earth, it could take a couple of
hours, maybe take a couple of days, but they would eventually
find it. With the avenue of AI, they have condensed it down to
minutes and hours. Within 6 months, they have condensed down
the amount of time that it takes to be able to find some
particular piece of equipment or object here on earth, so huge
advances.
I would never, ever, ever, ever want to see government
regulation, government requirements stifle our ability for this
American company to advance technology as fast as it has. It
has been remarkable, and I meet with this company about every 6
months, and the innovation that I see is tremendous. And there
are a lot of applications with a lot of different agencies, and
not just the public sector, not just DoD, but the private
sector as well. And so, any thoughts on, with this kind of
regulation happening, how we stay, at least not just one, two,
but a couple steps ahead of our adversaries around the world.
Mr. Thierer?
Mr. Thierer. Yes, very briefly, Congresswoman, you are
exactly right.
Ms. Mace. You got some time.
Mr. Thierer. But this is why I spoke in my testimony about
the symbiotic relationship between a strong technology base and
our national security interests because this is how we maintain
a strong security for the United States. Second, we should
point out that anything in this executive order that we are
discussing does not apply in China, and it does not apply to
these other nations I was just discussing----
Ms. Mace. That is right.
Mr. Thierer [continuing]. The UAE, Saudi Arabia, whatever
else.
Ms. Mace. They can do whatever they want.
Mr. Thierer. Whatever they want, right?
Ms. Mace. They do not have to follow U.S. law or
regulation.
Mr. Thierer. So, we cannot put our head into the sand and
think that just because we are constraining our companies, they
are constraining theirs.
Ms. Mace. Ms. Huddleston.
Ms. Huddleston. I just would like to add, I think the
example you provided shows why it is so difficult to define AI,
and why one of the concerns that I know Mr. Chilson expressed
about the definitions in the executive order that could reach
very far into everyday algorithms or everyday technologies that
we are using can be concerning, is in the defense context, we
often hear that mentioned as a high-risk scenario. But there
are many technologies, things that may be doing auto captions
for meetings or maybe helping to map certain areas that might
be useful in the defense context but are being developed for
these dual-use purposes. And we do not want to see a scenario
where they cannot be used by, say, the Department of Defense as
necessary, even though they are benign and beneficial
technologies.
Ms. Mace. Mr. Chilson?
Mr. Chilson. Well, we do have a template for how the last
disruptive technology, the last major disruptive sea change
technology of the internet was treated by the U.S. Government.
And there was a very specific choice, both in legislation and
at the executive level, to let the market lead, to let
innovators lead, to let them drive this technology forward
because they can explore a lot of different approaches and uses
in a way that we could never envision in the early 90s when
some of these decisions were made. And so, I do think that that
template, which requires action both by the White House, it can
require action by the White House, and also it could be done by
Congress as well, to take an open look at a structure that
would allow for a lot of variation. Let us focus on some
specific targeted harms but not tech specific. All the bad
things that we have heard about AI, those things are bad
whether or not they have been done with AI or with another
tool. We should target those harms, we should treat those
seriously, and we should deal with them, but there is no reason
to target AI on these specifically.
Ms. Mace. Dr. Lee?
Dr. Turner Lee. Yes, and I will just say, just with my
colleagues here, I think we can go back and look at the 5G
revolution, right, as an example of where we sort of stood back
on where we wanted to go as a Nation when it came to mobile
wireless leadership, right? And we eventually had to catch up
with China--I write a lot about that--and create our own
regulatory guardrails ourselves to ensure that not only were we
catching up with China, but we were also imparting in the
United States a type of social capital and economic capital and
innovation to actually expand those networks.
The same thing should be said about how we look at the AI
Act from the EU recently. I do not think anyone is really
saying in the EO that we need to come up with this broad
regulatory guidance that we all need to adhere to. I think what
we are seeing in the EO is here are some areas that we need to
take precaution with. Here are some areas where we need to
advance leadership, whether it is in the workforce or whether
it is in innovation, whether it is in the adjacent products and
services that go with the supply chain.
At the end of the day, we are the United States, right, and
we are going to do things a little differently than everybody
else, but without making this a Wild West when it comes to
innovation and having some certainty that redeposits back into
our American economy, you know, we are going to find ourselves
in the same similar situations we have had with other
technology disruptions.
Ms. Mace. OK. Thank you all for your questions, and I will
now yield to Mr. Lynch for 5 minutes.
Mr. Lynch. Thank you very much, Madam Chair, and thank you
for your relentless leadership on this issue. Much appreciated.
I had a couple of hearings at the same time. That is how they
do everything here. Everything happens at once, but I
apologize. I was in a hearing down the corridor on Financial
Services. I want to thank the witnesses. I have read your bios,
and I appreciate the expertise and the intellect that you bring
to this hearing. So, thank you very much for helping the
Committee with its work. This is one of those areas that we are
really grappling with. The velocity of change has been
incredible, and I think we are racing to catch up.
As reported by the United States intelligence community in
its Worldwide Threat Assessment recently, Russia, China, North
Korea, and other state actors are continuing to conduct malign
influence operations aimed at disseminating disinformation and
magnifying U.S. societal divisions and interfering with the
upcoming U.S. elections. They are doing this in other countries
as well, but we are principally concerned about this country
and this election right now. So, as you might imagine, the
introduction of generative AI sort of amplifies the
possibilities of this exacerbated threat.
I know that FBI Director Chris Wray recently testified to
the Select Committee on the Chinese Communist Party. He said
that, ``This election cycle, the U.S. will face more
adversaries moving at a faster pace and enabled by new
technology,'' speaking of AI. He also went on to say that
advances in generative AI make it ``easier for both more and
less sophisticated foreign adversaries to engage in malign
influence.''
And today, virtually, as you know, anyone can weaponize AI
to create fake but convincing photos, videos, and audio clips
with the purpose of election interference or manipulation. Just
recently, 2 days before the New Hampshire Presidential
primary--I am sure you heard of it--thousands of residents of
New Hampshire received robocalls that used an AI voice cloning
software to imitate President Biden. It was rather convincing
as well. The robocall encouraged recipients not to vote in the
primary election and ``save your vote for the November
election,'' so you can easily see how, you know, this
technology might be used for nefarious purposes.
Dr. Turner Lee, what effect can we expect AI-generated
deepfakes and misinformation to have on Americans' trust, their
trust in the election process?
Dr. Turner Lee. Thank you for that, and I am very happy
that you brought this up as part of this conversation. As the
Brennan Center has reported on a variety of their reports, we
definitely need to address AI-generated content that is leading
into misinformation and disinformation to ensure that we have
an informed electorate and to also have electoral
infrastructure in process that is not harmed by malicious
actors or other malfeasances that are relying on synthetic
media or artificially generated content that will dissuade
voters. And that, I think, applies to everybody in this room,
regardless of your party, your residents, et cetera. You know,
we all need to make sure we are going to the polls with that
kind of information.
With that being said, it is really important for us to get
a handle on this. I think one of the things that we did not do
in this panel that we often do in others is we sort of talk
about AI and generative AI as if they are the same thing. You
know, generative AI has more advanced capabilities of
extraction of voice, image, text in ways that we cannot often
find out where it originated from. And so having, as we see in
the executive order, these industries commit to helping us with
a better digital watermarking system, being able to have
conversations around copyright protections, really determining
ways in which we engage the public in general education so that
they are more informed about misinformation, I think are
particularly important. And Congress has actually had some
bipartisan action on this that I think we should take heed of
if we are going to get these elections right now and into the
future.
I would just also suggest to you, Congressman, that states
are also taking this very seriously. My colleagues at
Brookings, we are looking at, you know, what states are doing,
and I think there are some synergies between Federal action and
concern and state action and concern. And so again, broadening
the scope of making sure that this is a priority for Congress,
I think, is key.
Mr. Lynch. Thank you. I just want to ask, are there
countermeasures that are at hand to allow us to sort of push
back on some of this and reveal its, you know, I guess,
negative nature and its falsity?
Mr. Thierer. Yes. I will just mention one----
Mr. Lynch. Sure.
Mr. Thierer [continuing]. And my colleagues will have more.
Representative Rochester has a really good bill having to do
with AI literacy and education, and trying to find ways to
teach our electorate and our citizens that there are dangers
out there, including misinformation in campaigns and elsewhere
in the market. So, that is a good baby step to take to
partially get at this problem, which is a serious one.
Mr. Lynch. I was thinking more about technology that could
vet, you know, to look at. I know it is incredibly difficult,
and the technology is changing so quickly, but are there proven
methods that might allow us to uncover quickly, you know, a
message that is not from its proposed source?
Ms. Huddleston. We have seen the industry start to evolve
to respond to these concerns, and there will be different
actions from different players, just as there are different
natures of what exactly AI looks like. So, how a certain social
media platform may respond to concerns about AI-generated
images in an election context or in an election ad context may
look different than how a search engine does or how another
tool does when it is, for example, dealing with voice as
opposed to video. We have seen that, oftentimes, allowing these
different norms to play out will allow consumers to get
appropriate amounts of information in that context because it
will look different on each platform and each tool, as opposed
to a government one-size-fits-all approach, where not only do
you have concerns about the potential impact on speech, you
also have concerns about, given the broad definition of AI, how
much content could this apply to, and might it bring in things
that are more common. So, for example, something that removes
an object in the background using an AI editing tool but gets
labeled as AI generated or AI manipulated because it used an AI
tool rather than a human graphic designer or something.
Mr. Lynch. Thank you. Madam Chair, I appreciate the
courtesy you have extended me, and I will yield back because I
know I am way over time.
Ms. Mace. No, you are good. Thank you, Mr. Lynch, and in
closing, I want to thank our panelists for being with us today
and providing their testimony and answering our questions. We
want to make sure that the United States is the clear winner,
the clear innovator in all technology, including AI. We do not
want to stifle innovation. We do not want to stifle investment
in AI or its innovation in any way, shape, or form. There is a
lot at risk here, and we do not want our adversaries getting
ahead of us or giving them the room or the rope to get ahead of
us at all, so thank you again.
And with that, and without objection, all Members will have
5 legislative days within which to submit materials and to
submit additional written questions for our witnesses, which
will then be forwarded to the witnesses for their response.
If there is no further business, without objection, the
Subcommittee stands adjourned.
[Whereupon, at 11:30 a.m., the Subcommittee was adjourned.]
[all]