[Senate Hearing 118-223]
[From the U.S. Government Publishing Office]


                                                         S. Hrg. 118-223

                   THE USE OF ARTIFICIAL INTELLIGENCE
                      AT THE LIBRARY OF CONGRESS,
                     GOVERNMENT PUBLISHING OFFICE,
                      AND SMITHSONIAN INSTITUTION

=======================================================================

                                HEARING

                              BEFORE THE

                 COMMITTEE ON RULES AND ADMINISTRATION
                          UNITED STATES SENATE

                    ONE HUNDRED EIGHTEENTH CONGRESS

                             SECOND SESSION

                               __________

                      WEDNESDAY, JANUARY 24, 2024

                               __________

    Printed for the use of the Committee on Rules and Administration
    
[GRAPHIC NOT AVAILABLE IN TIFF FORMAT]    


                  Available on http://www.govinfo.gov
                  
                              __________

                   U.S. GOVERNMENT PUBLISHING OFFICE                    
54-677                      WASHINGTON : 2024                    
          
-----------------------------------------------------------------------------------                     

                 COMMITTEE ON RULES AND ADMINISTRATION

                             SECOND SESSION

                  AMY KLOBUCHAR, Minnesota, Chairwoman

CHARLES E. SCHUMER, New York         DEB FISCHER, Nebraska
MARK R. WARNER, Virginia             MITCH McCONNELL, Kentucky
JEFF MERKLEY, Oregon                 TED CRUZ, Texas
ALEX PADILLA, California             SHELLEY MOORE CAPITO, West 
JON OSSOFF, Georgia                      Virginia
MICHAEL F. BENNET, Colorado          ROGER WICKER, Mississippi
PETER WELCH, Vermont                 CINDY HYDE-SMITH, Mississippi
LAPHONZA R. BUTLER, California       BILL HAGERTY, Tennessee
                                     KATIE BOYD BRITT, Alabama

                    Elizabeth Farrar, Staff Director
                Jackie Barber, Republican Staff Director


                         C  O  N  T  E  N  T  S

                              ----------                              
                                                                  Pages

                         Opening Statement of:

Hon. Amy Klobuchar, Chairwoman, a United States Senator from the 
  State of Minnesota.............................................     1
Hon. Deb Fischer, a United States Senator from the State of 
  Nebraska.......................................................     3
Dr. Carla Hayden, Librarian of Congress, Library of Congress, 
  Washington, DC.................................................     4
Hon. Hugh Nathanial Halpern, Director, United States Government 
  Publishing Office, Washington, DC..............................     6
Meroe Park, Deputy Secretary and Chief Operating Officer, 
  Smithsonian Institution, Washington, DC........................     8

                         Prepared Statement of:

Dr. Carla Hayden, Librarian of Congress, Library of Congress, 
  Washington, DC.................................................    19
Hon. Hugh Nathanial Halpern, Director, United States Government 
  Publishing Office, Washington, DC..............................    26
Meroe Park, Deputy Secretary and Chief Operating Officer, 
  Smithsonian Institution, Washington, DC........................    33

                            For the Record:

Testimony for the Record Submitted by Marci Harris, POPVOX 
  Foundation.....................................................    35

                  Questions Submitted for the Record:

Hon. Amy Klobuchar, Chairwoman, a United States Senator from the 
  State of Minnesota to Dr. Carla Hayden, Librarian of Congress, 
  Library of Congress, Washington, DC............................    40
Hon. Deb Fischer, a United States Senator from the State of 
  Nebraska to Dr. Carla Hayden, Librarian of Congress, Library of 
  Congress, Washington, DC.......................................    42
Hon. Shelley Moore Capito, a United States Senator from the State 
  of West Virginia to Dr. Carla Hayden, Librarian of Congress, 
  Library of Congress, Washington, DC............................    47
Hon. Amy Klobuchar, Chairwoman, a United States Senator from the 
  State of Minnesota to Hon. Hugh Nathanial Halpern, Director, 
  United States Government Publishing Office, Washington, DC.....    52
Hon. Deb Fischer, a United States Senator from the State of 
  Nebraska to Hon. Hugh Nathanial Halpern, Director, United 
  States Government Publishing Office, Washington, DC............    54
Hon. Shelley Moore Capito, a United States Senator from the State 
  of West Virginia to Hon. Hugh Nathanial Halpern, Director, 
  United States Government Publishing Office, Washington, DC.....    57
Hon. Amy Klobuchar, Chairwoman, a United States Senator from the 
  State of Minnesota to Meroe Park, Deputy Secretary and Chief 
  Operating Officer, Smithsonian Institution, Washington, DC.....    58
Hon. Deb Fischer, a United States Senator from the State of 
  Nebraska to Meroe Park, Deputy Secretary and Chief Operating 
  Officer, Smithsonian Institution, Washington, DC...............    59
Hon. Shelley Moore Capito, a United States Senator from the State 
  of West Virginia to Meroe Park, Deputy Secretary and Chief 
  Operating Officer, Smithsonian Institution, Washington, DC.....    62

 
                   THE USE OF ARTIFICIAL INTELLIGENCE
                      AT THE LIBRARY OF CONGRESS,
                     GOVERNMENT PUBLISHING OFFICE,
                      AND SMITHSONIAN INSTITUTION

                              ----------                              


                      WEDNESDAY, JANUARY 24, 2024

                               United States Senate
                      Committee on Rules and Administration
                                                    Washington, DC.
    The Committee met, pursuant to notice, at 2:15 p.m., in 
Room 301, Russell Senate Office Building, Hon. Amy Klobuchar, 
Chairwoman of the Committee, presiding.
    Present: Senators Klobuchar, Fischer, Padilla, Ossoff, 
Capito, Hagerty, and Britt.

         OPENING STATEMENT OF HONORABLE AMY KLOBUCHAR,

            CHAIRWOMAN, A UNITED STATES SENATOR FROM

                     THE STATE OF MINNESOTA

    Chairwoman Klobuchar. Calling this hearing to order. Good 
afternoon. I would like to thank Ranking Member Fischer, as 
always, and our colleagues for helping us put together this 
hearing. With us today is Dr. Carla Hayden, who is the 
Librarian of Congress. Thank you. We always enjoy having you 
here.
    Also here is Hugh Halpern, the Director of the Government 
Publishing Office. Back again, thank you. As well as Meroe 
Park, thank you for being here, the Deputy Secretary and Chief 
Operating Officer of the Smithsonian. We understand that the 
Secretary is ill. Just--he will recover but was not able to 
make it, so we really appreciate it.
    Today, we are going to talk about the very important topic 
of artificial intelligence and the agencies that play such a 
critical role in serving the American people. AI has the 
potential, as we know, to lead to incredible innovation by 
supercharging scientific research, improving access to 
information, and increasing productivity. But like any emerging 
technology, AI comes with significant risks and our laws need 
to be as sophisticated, as the potential threats are there to 
our own democracy.
    Understanding these risks and benefits has been a major 
bipartisan focus of the Senate, with Senators Schumer, Rounds, 
Young, and Heinrich leading a series of nine forums since the 
fall, and a number of us on various committees working on 
proposals so that we are going to be ready when this hits.
    I think, as you all saw from the reports of the robocalls 
in New Hampshire with the fake voice of the President or other 
things that

[[Page 2]]

have also occurred to candidates on the Republican side of the 
aisle, this is not going to be one side or another. It is 
something that we as a Congress have to deal with to put some 
guardrails. That includes, of course, the work that goes on in 
these agencies.
    At our hearing in September, all the witnesses agreed that 
AI poses risks to our elections, and we heard testimony on why 
we must work to put guardrails in place. That is why I am 
leading a bipartisan bill with Senators Hawley, Coons, Collins, 
and also joined with Senators Bennet and Ricketts, to prohibit 
fraudulent AI-generated content in our elections within the 
framework of the Constitution, for instance, allowing for 
satire and the like, and why we need to take other steps like 
disclaimers on ads that use AI so that the citizens of this 
country can actually believe that it is their own candidate or 
their opposing candidate who is speaking.
    Another example of legislation going on. Senator Thune and 
I have joined together in introducing a bill to put in place 
commonsense safeguards for the highest risk, non-defense 
applications of AI, and improve transparency. I see that 
Senator Capito is here. She has also joined us on this 
important bill that is mostly coming out of the Commerce 
Committee.
    Let us get to your stuff, AI and how it affects the work of 
the three agencies before us, the Library of Congress, 
Government Publishing Office, and the Smithsonian. While it is 
important that our three witnesses today speak to measures they 
are taking to safeguard against potential harms, they are also, 
I think it is important to note, using AI technology in their 
work to protect our country's greatest treasures, advance 
scientific research, and improve public access to information.
    For example, the Library of Congress is testing emerging AI 
technology to expand how researchers can better use the 
resources they already house in their collections, which make 
up the largest library in the world, we note with much 
humbleness, such as a new AI tool that lets users instantly 
search through 1.56 million images from digitized historical 
newspapers to assist in archival research.
    GPO is working to harness the efficiencies offered by AI to 
modernize how it makes information from all three branches of 
government more usable for the public since much of its work 
has expanded to digital publishing, hence their new name, 
printing to publishing.
    As part of its work producing government documents, GPO is 
using AI to ensure quality control of items such as the 
material you use to print passports. It printed 22 million of 
them, as we learned at the last hearing last year.
    Finally, at the Smithsonian, which is the world's largest 
museum, education, and research complex, researchers are 
exploring how to use AI to do things like tackle some of the 
most challenging problems in astrophysics, classify species of 
fish in the Amazon, and make collections more accessible, 
accurately identifying the contributions of women, I like this 
one, in historical texts, in which they were often identified 
in writing by only their husbands' names. That will be 
interesting what you discover with AI.
    We must continue working to stay ahead of the curve, and I 
am committed to working in a bipartisan way with Senator 
Fischer so that our country can benefit, and your agencies can 
benefit from

[[Page 3]]

the best of AI, while protecting against any threats. Thank you 
again, and I will turn it over to Senator Fischer.

  OPENING STATEMENT OF HONORABLE DEB FISCHER, A UNITED STATES 
               SENATOR FROM THE STATE OF NEBRASKA

    Senator Fischer. Thank you, Chairwoman Klobuchar, for 
calling this hearing on The use of Artificial Intelligence at 
the Library of Congress, the Government Publishing Office, and 
the Smithsonian. I also want to thank our three witnesses who 
are here with us today.
    I know we are all wishing Secretary Bunch a speedy 
recovery. I believe this is the first time the Committee has 
heard from the Library of Congress, the Government Publishing 
Office, and the Smithsonian at the same time, and Director 
Halpern, it is nice to see you back with us so soon.
    The Library of Congress and the Smithsonian both safeguard 
a vast collection of our Nation's treasures, welcome our 
constituents to their beautiful buildings to learn and explore, 
and serve as crucial resources to the scientific and the 
academic communities.
    The Government Publishing Office also performs vital 
functions for the American people as it produces, distributes, 
preserves, and publishes documents for all three branches of 
our government.
    This is the Rules Committee's second hearing on the use of 
AI. As we discussed at the first hearing, there is no question 
that AI is transformative, and it is poised to evolve rapidly. 
While AI brings the possibility of creating efficiencies and 
competitive advantages across government, it also creates 
risks.
    Understanding and weighing the benefits and risks of AI are 
necessary first steps before widely adopting the use of AI. 
Today, I look forward to hearing from our witnesses about how 
our legislative branch agencies are exploring potential 
applications of AI.
    There are many important questions to ask, such as are 
these existing issues or specific challenges at our legislative 
branch agencies that AI could solve? Are there existing 
services that our agencies provide that AI could improve? Can 
AI provide new avenues for these agencies so that you can 
fulfill your missions?
    I also look forward to hearing about the guardrails our 
agencies are putting in place to mitigate those risks of AI. We 
need to understand the cost of AI and AI's impacts on agencies' 
budgets.
    Just as importantly, agencies need to understand the 
implications of AI use, especially the risks related to 
privacy. Understanding legislative branch agencies' use of AI 
is an important oversight question for this Committee. That 
said, we cannot lose sight of existing challenges facing the 
agencies that we oversee.
    As the Library looks to hire a new director for the 
Congressional Research Service, and GPO seeks to maintain high 
quality staffing standards in the face of your retiring work 
force, it is important that this Committee supports their 
efforts to improve hiring and retention efforts.
    Furthermore, as the Smithsonian endeavors to establish two 
new museums authorized by statute, we must maintain rigorous 
oversight of their efforts to tackle a deferred maintenance 
backlog, a chronic issue for that agency.

[[Page 4]]

    With that said, I look forward to a productive discussion 
about our legislative branch agencies' use of AI, as well as 
other issues that you face. Thank you, Madam Chair.
    Chairwoman Klobuchar. Well, thank you so much, Senator 
Fischer. I will introduce our witnesses. Our first witness, as 
I noted, is Librarian of Congress, Dr. Carla Hayden.
    Dr. Hayden was sworn in in September 2016. She has done a 
tremendous job leading the Library. She previously served as 
the CEO of Enoch Pratt Free Library in Baltimore, and she 
received her undergraduate degree from Roosevelt University, 
and master's and Ph.D. from the University of Chicago, that I 
also attended.
    Whenever I see Dr. Hayden, I always think about Barbara 
Mikulski, who was so proud that a Baltimore librarian got this 
job, and you have made her proud too. Thank you.
    Our second witness is Director of the Government Publishing 
Office, Hugh Halpern, who has been in his position since his 
unanimous confirmation, it does not happen all the time, by the 
Senate in December 2019.
    He previously worked on Capitol Hill for more than 30 
years, including for former Speaker Paul Ryan. He received his 
bachelor's and master's degrees from American University, and 
his law degree from George Mason. Thank you, and we really 
enjoyed your attendance at your last hearing.
    Our final witness is Meroe Park, who became Deputy 
Secretary and Chief Operating Officer of the Smithsonian in 
December 2019. Previously, she worked for 27 years at the CIA. 
Okay, I did not know that. When I read these bios, I learn new 
things all the time. You probably do not advertise that, but 
that is interesting. Including as executive director and chief 
operating officer, and she received her undergraduate degree 
from Georgetown.
    Okay, now we are going to swear in the witnesses. Okay, if 
you raise your right hand, do you swear that the testimony you 
will give before the Committee shall be the truth, the whole 
truth, and nothing but the truth, so help you God?
    Dr. Hayden. I do.
    Mr. Halpern. I do.
    Ms. Park. I do.
    Chairwoman Klobuchar. Thank you, and you can be seated. We 
will now proceed to five minute testimonies. We will start with 
Dr. Hayden.

           OPENING STATEMENT OF DOCTOR CARLA HAYDEN,

   LIBRARIAN OF CONGRESS, LIBRARY OF CONGRESS, WASHINGTON, DC

    Dr. Hayden. Thank you, Madam Chairwoman, Ranking Member 
Fischer, and Members of the Committee. Thank you for this 
invitation to appear before you to discuss a topic that is of 
great interest across the government and within the public 
archives sector, artificial intelligence.
    I am pleased to have the opportunity to further engage with 
my legislative branch colleagues on this topic and to update 
the Committee on the Library's activities exploring AI. At the 
Library, there are four areas of focus with AI, to expand 
access to our collections, to enhance services for users, to 
improve internal processes

[[Page 5]]

for increased efficiency, and to implement a slate of strong 
governance for the use of AI.
    Becoming a more digitally enabled agency has been a key 
focus of my tenure as Librarian. In 2019, the Library published 
a comprehensive digital strategy to guide the agency's use of 
technology in an increasingly digital world. Building on this 
first major step, we have now fully integrated our digital 
strategy into our 2024-2028 strategic plan, and the plan 
embraces a central idea, that technology must be, as our CIO, 
Judith Conklin says, baked into all we do. Since 2018, the 
Library's Digital Innovation Division, also affectionately 
known as LC Labs, has investigated AI and shared the results of 
its research experiments with the public.
    Working with digital researchers we call Innovators in 
Residence, LC Labs has made its mark as a launch pad for 
innovating engaging uses of AI to expand access to our 
collections. A very popular example is Citizen DJ, a music 
sampling application that allows users to remix and create 
music using free-to-use non rights restricted audio from the 
Library's collections.
    Other exciting AI use cases include experimenting with 
machine learning and optical character recognition, or OCR, to 
help manage metadata and machine readable text for digitized 
documents.
    For example, OCR has increased the discoverability of more 
than 20 million historic American newspaper pages through the 
Chronicling America Project. Building on that technology, users 
can also search those historic newspapers for photos using an 
application LC Labs rolled out in 2020 called Newspaper 
Navigator.
    Enhancing services to Congress and the public is also a 
major area of focus. In 2022, we released a Congress.gov API, 
application programing interface, to make it easier for the 
public to access and use accurate, structured Congressional 
Legislative data.
    Additionally, several Library service units are 
successfully demonstrating AI's use in bolstering the Library's 
information services. Our digital innovators have been working 
with the Copyright Office to test approaches for extracting 
data from historical copyright records.
    This project combines human skills with AI capacity to make 
a written or handwritten analog record more accessible and 
easier to search online, and it is just one example of how 
humans in the loop can be the model for AI and successful 
integration of technology with human skill, because a very 
important principle is that well trained human beings will 
always be critical to the work of the Library of Congress.
    Now, the Library's AI use cases have also been testing 
opportunities to increase efficiency and staff productivity. 
For example, our Cataloging Division is currently experimenting 
with AI to help staff more efficiently process bibliographic 
information and catalog records.
    There are many opportunities with AI, but there also must 
be robust governance. The Library is in a strong position as 
our existing technology governance policies provide an 
adaptable foundation to use in guiding us with emerging 
technologies like AI. Further, the Library's approach to 
implementing AI closely aligns with the practices of other 
federal institutions and is informed by NIST's AI Risk

[[Page 6]]

Management Framework, which also aims to improve the 
trustworthiness of AI applications.
    Responding to this fast developing area of technology calls 
for collaboration across the private sector and government, 
which is why the Library participates in the General Services 
Administration AI Community of Practice and is a leading member 
of the International Artificial Intelligence for Libraries, 
Archives and Museums Secretariat.
    To conclude, discovering the role AI has to play in 
enhancing services to Congress and other users remains an 
ongoing effort. As the Library of Congress charts this course 
forward, we plan to draw on our history of technological 
innovations, our rigorous development of standards, and the 
input of our stakeholders and partners to align possible uses 
for the public.
    We appreciate this opportunity, and we hope that you will 
join with us in making sure the values of transparency, 
accountability, and efficacy remain as we explore this 
opportunity.
    [The prepared statement of Dr. Hayden was submitted for the 
record.]
    Chairwoman Klobuchar. Thank you. Mr. Halpern.

         OPENING STATEMENT OF HONORABLE HUGH NATHANIAL

HALPERN, DIRECTOR, UNITED STATES GOVERNMENT PUBLISHING OFFICE, 
                         WASHINGTON, DC

    Mr. Halpern. Thank you. Good afternoon, Chairwoman 
Klobuchar, Ranking Member Fischer, Senator Capito. It is always 
good to see you. I am pleased to appear before you today to 
share some of the potential uses of AI and related technologies 
at the Government Publishing Office.
    GPO differs from the Library and the Smithsonian in that it 
is fundamentally a manufacturing operation. We publish, 
produce, and maintain materials for all three branches of 
government.
    Our 1,600 craftspeople and professionals produce virtually 
all of Congress's documents, along with numerous other 
publications, and manufacture secure credential products like 
the United States passport. We also provide digital information 
either through our own trusted digital repository, Govinfo.gov, 
or by serving data to our partners like the Library where they 
use that data on sites like Congress.gov.
    No matter what you call it, artificial intelligence, 
machine learning, or a large language model, GPO's operations 
are just as susceptible to disruption as any commercial firms, 
and that is not necessarily a bad thing.
    My written testimony describes GPO's policy approach to 
this new generation of tools, so my statement today will focus 
on three potential applications for AI and related technologies 
in our day-to-day operations.
    First, we believe these tools can improve our quality 
assurance process by automatically recognizing defects that a 
human inspector might miss. We already use a rudimentary form 
of this technology in the production of the current version of 
the United States passport. GPO uses equipment that optically 
scans the pages that will become the identity page in a 
personalized passport.

[[Page 7]]

    This equipment looks at each strip of three pages for 
variances that exceed the specifications for the material and 
rejects those pages that do not conform to the standard. AI 
technology has the potential to further refine this review, 
allowing the machines to learn what may constitute a natural 
variation that is within the specification and what is not.
    This has the potential to reduce defect rates, lower waste, 
and free up our quality assurance team to focus on solving 
bigger quality problems as they arise.
    Second, we see tremendous potential for supplementing our 
proofreading team. Proofreaders are very difficult to hire, and 
we need to free them from making routine, repetitive 
corrections and allow them to focus on more subtle issues that 
really require a human being to interpret.
    One example is capitalization. GPO style says that we 
capitalize the letter ``s'' in the word ``state'' when we are 
referring to a political subdivision of the United States. 
Currently, we use computer scripts to perform global search and 
replace functions on documents to correct our most common 
errors, including that one. But those scripts are really blunt 
instruments.
    For instance, they cannot tell the difference between the 
``State of Minnesota'' and a ``New York state of mind.'' AI 
holds the promise of tools that understand context and know 
when text refers to one kind of state or the other. That will 
cut down on the need for our proofreaders to review and re-
review, and then correct material that has already been run 
through our automated tools, and free them up to focus on more 
difficult contextual issues.
    My final example comes from GPO's public information 
mission. We have had great success in making congressionally 
mandated reports publicly available since the end of last year, 
with more than 180 of those reports available on GovInfo as of 
yesterday.
    Most of those reports come to us as PDFs. While that is a 
good format to show how the printed document looks, it is not 
always the best format for viewing on a phone, a tablet, or 
even for folks with disabilities.
    While GPO would like to get these reports in more flexible 
formats like XML, the agencies authoring reports are not always 
equipped to supply them that way. AI technologies hold the 
promise of allowing us to extract the information from a PDF, 
understand the document structure, and produce an alternative 
view that works on different kinds of devices, all without 
manual, time consuming work from our team.
    These are just three examples where we see applications in 
GPO's operations, and we are considering many pilots in the 
future. All of these are intended to act as a force multiplier 
for our team, allowing our folks to be more productive, and 
deliver a higher value for our customers and the taxpayer.
    Madam Chairwoman, Ranking Member Fischer, thank you for the 
opportunity to testify before the Committee today. I look 
forward to any questions the Committee may have.
    [The prepared statement of Mr. Halpern was submitted for 
the record.]
    Chairwoman Klobuchar. Thank you very much. Last but not 
least, Ms. Park, thank you.

[[Page 8]]


       OPENING STATEMENT OF MEROE PARK, DEPUTY SECRETARY
     AND CHIEF OPERATING OFFICER, SMITHSONIAN INSTITUTION, 
                         WASHINGTON, DC

    Ms. Park. Thank you. Thank you for the opportunity to 
present today. I will pass along your well wishes to the 
Secretary. Early in our time as an institution, we collected 
meteorological data from volunteers nationwide via telegraph.
    Fast forward to a few years ago, when we helped create the 
first image of a black hole at the center of the Milky Way. In 
short, we have always embraced revolutionary technology, from 
the locomotive to the pocket watch, from the telegraph to the 
internet, and artificial intelligence is no different.
    Throughout the institution, scientific uses of machine 
learning are actually quite common, applying it to the 
identification of exoplanets, tracing the evolutionary history 
of pollen, and conducting field research in conservation. In 
other words, artificial intelligence has been around the 
Smithsonian for almost a decade. What has changed is the 
widespread availability of generative AI.
    These widely available applications of AI hold great 
promise, from making us more efficient and effective as an 
organization, to improving the visitor experience and extending 
our reach across the Nation. We are learning from our current 
use cases and are exploring new ones as this technology 
continues to evolve.
    For example, the National Museum of African American 
History and Culture's Freedmen's Bureau Transcription Project, 
which is the Smithsonian's largest crowd sourcing initiative, 
is transcribing genealogical records of the formerly enslaved. 
In a recent trial run, AI tools reliably transcribed nearly two 
thirds of the test content, showing how AI has the ability to 
dramatically scale up work with large data sets.
    Other examples, including, as the Chairwoman mentioned 
during her opening remarks, our data science labs' development 
of a model that can discover incorrect women's contributions 
mistakenly attributed to men in our collections.
    The Smithsonian Astrophysical Observatories Astro AI Center 
has more than 50 AI projects planned or underway, creating 
excitement about our solar system and the universe itself.
    Inside the Smithsonian, we recently announced the creation 
of an AI Community of Practice, which will help us test new 
capabilities across the institution, as well as build a 
framework of good governance to determine the best way to use 
AI effectively and responsibly.
    People rely on our reputation as a trustworthy reservoir of 
knowledge, so we understand concerns about bias, ethics, 
safety, and accuracy in available AI outcomes, but there are 
also fantastic opportunities as well. We and other cultural 
institutions can collaborate with technology leaders to help 
improve AI tools for everyone.
    We can also examine the many dimensions of AI from a 
scholarly perspective, convening experts to examine its 
ethical, social, and economic implications. The nation's 250th 
birthday in 2026 will be a pivotal moment for us to experiment 
with innovative AI tools, to enable collaboration with tech 
leaders in other government and cul

[[Page 9]]

tural organizations to make AI more reliable and trustworthy, 
to educate people about the benefits and the risks of AI.
    Ultimately, AI is a tool. Used properly, it will allow the 
Smithsonian to expand our reach and our impact. Like any new 
technology, it comes with risks and the potential for 
unintended consequences. But by developing internal expertise 
and partnering with external experts, we can thoughtfully 
leverage AI, both for the Smithsonian and for the American 
public. Thank you.
    [The prepared statement of Ms. Park was submitted for the 
record.]
    Chairwoman Klobuchar. Thank you very much. I am going to 
turn it over to Senator Fischer and--for her questions, and 
then Senator Padilla.
    Senator Fischer. Thank you, Madam Chairwoman. Thank you all 
once again for being here today. Can each of you share whether 
your respective agencies have an official policy that governs 
its use of AI? Dr. Hayden, do you have an official policy on 
that?
    Dr. Hayden. Yes. The Library has a number of policies for 
security, IT security in general, and we have an AI working 
group that has developed guidelines.
    We also follow the NIST's framework. We also are very 
cognizant and use the Executive Orders that have also laid that 
foundation. We are very, I would say zealous on making sure 
that we have those frameworks and guidelines.
    Senator Fischer. You mentioned in your oral testimony, it 
is in your written as well, about NIST and it aims to improve 
the trustworthiness of AI applications. Can you give me an 
example of how that would be used? I think a lot of us, when we 
think of AI, we do not get down to the nitty gritty.
    We do not get down to the basics of what it really can do. 
We just think, oh, this is going to be so wonderful. It will be 
sorting data. I think this--I think some of your answers from 
all of you would be very helpful on that.
    Dr. Hayden. One of the through lines for each of us is 
trustworthiness and with the Library of Congress, we take very 
seriously being stewards of historic information and current 
information so that we are a trusted source.
    Working with the framework that includes looking at a new 
technology, you map it, you explore, and then you measure its 
usefulness and effectiveness. Then if it passes quite a few of 
these things, you would look at how you could implement it.
    Being able to make sure and having these templates, we have 
questions that we pose when anything is being proposed. But 
with AI in particular, we ask what does it do? How can we make 
sure that humans are in the loop for their quality control? It 
is a very important part of looking at AI in particular and any 
newer technology.
    Senator Fischer. Thank you. Mr. Halpern, can you let me 
know if you--do you have a policy?
    Mr. Halpern. Absolutely. We have a directive in place since 
last month, since the end of 2023. One of the things that that 
created was our AI Governance Committee, and I believe that 
committee is going to have their first kickoff meeting on 
Monday, where they are going to start evaluating different 
technologies for some pilot

[[Page 10]]

deployments in the near term so we can start testing some of 
these in our environment.
    Senator Fischer. You know, you mentioned in your testimony, 
a lot of--the three interesting examples you gave of how you 
are using AI right now.
    How do you decide on a case-by-case basis when you are 
ready to test that on that project, whether you think it is 
safe at that point in time, or if you are just trying to gather 
information?
    Mr. Halpern. We already have a very robust system for 
evaluating software and hardware for deployment at GPO.
    Much of what we do is mission critical to the Congress, so 
we need to make sure that anything we deploy does not interfere 
with our ability to deliver for our customers, whether that is 
the House or the Senate, or our State Department customer, or 
any other customer.
    This Governance Committee is another layer on top of that. 
Looking at some of these new technologies, both to see if there 
are improvements to the way we do things that can deliver 
benefits to our customers and minimize the risk from those same 
technologies of introducing variables into our product. For 
instance----
    Senator Fischer. What does that mean, the risks of 
introducing variables--?
    Mr. Halpern. We have all read about the issues with ChatGPT 
hallucinating facts that are not there. The good news is GPO 
does not generate a lot of public information. We make other 
people's public information available.
    Much of where we can leverage these technologies is really 
in the quality assurance world. But, for the same reason you 
need to double-check ChatGPT's output, we need to make sure 
that the optical system that is checking the quality of our 
passport identity pages is not introducing or missing quality 
problems.
    From that perspective, we really need to deploy these 
systems in a limited way so that we can make sure they are 
doing what they are advertised to do so that we are not 
delivering a product to our customer that is of lesser quality 
than it should be.
    Senator Fischer. Can I have the Smithsonian response as 
well.
    Ms. Park. Of course.
    Senator Fischer. Thank you. Ms. Park.
    Ms. Park. Great. Thank you. I will address maybe two 
aspects of your question. Internally at the Smithsonian, we 
absolutely are focused on governance and also the sharing of 
best practices.
    This community of practice that we have formed is really 
designed to do a number of things. One is to identify policies 
that we already have that need to be updated with what is a 
very quickly evolving technology. Also, things like protecting 
privacy.
    You asked about specific examples of how does this actually 
play out in an organization. You have to make sure that if 
employees are using ChatGPT or some other generative AI tool, 
that we are thinking about how that data is being prepared and 
made available to AI to analyze.
    There is a lot of thought going into how we go about doing 
that internally. Then externally, we also have a public 
mandate, as several of my colleagues have talked about here, as 
trusted sources for the American public.

[[Page 11]]

    We want to make sure that we are available to them, that we 
have our 177 years of experience with history, culture, 
science, education as a way to inform these tools, and also as 
a way to help evaluate the accuracy of information.
    So that is still aspirational for us at this point, but we 
are thinking about both aspects of our internal governance, our 
internal opportunities for efficiency, as well as our ability 
to be a convener and to help in the more public realm.
    Senator Fischer. Thank you.
    Chairwoman Klobuchar. Okay, Senator Padilla. Thank you.
    Senator Padilla. Thank you, Madam Chair. As I have had a 
chance to share in this Committee before--before entering the 
Senate, I served as California's Secretary of State, where a 
part of the portfolio was serving as a custodian of the state's 
archives.
    At the time, I made it a priority to dramatically increase 
public access to our state's archives, and this involved a 
significant undertaking to digitize our archive materials to 
make them easily viewable, not just to the people of 
California, but to the world.
    Question, first question is for Ms. Park. You noted that 
prioritizing digitization is one of the Smithsonian's key 
considerations with respect to AI. How do you envision 
leveraging AI to increase public access to the Smithsonian?
    Ms. Park. Indeed. I will maybe make a couple of comments in 
response to your question. The first is that digitization is 
absolutely one of our priorities as an institution. We have 
been at it actually for a while. It is a behemoth task with 
over 150 million objects, and we are working hard to identify 
ways to prioritize that. We have a long ways to go in terms of 
having those materials digitized.
    When you refer to your own experience in California, the 
first step, obviously, with AI is even getting the records 
available and readable and analyzable in a digital format. So 
that is job number one for us.
    The second part, is we believe that, unlike a search engine 
that would be able to help the public discover certain items 
that we might have in our collection and learn about them, AI 
has the ability for the public to make connections among the 
documents and objects and items that we have in ways that we 
cannot do as humans.
    That we can use generative AI, and we hope to make it 
available to the public as well, to be able to discover things 
that we did not even realize we had in our collections. I think 
there is some really exciting opportunities here, but as you 
noted, job one is getting more of our items digitized and 
readable in a way that AI can make use.
    Senator Padilla. I will let you in on a secret, maybe a 
good strategy. Whatever is most popular already, that is where 
to start.
    Ms. Park. The ruby slippers.
    Senator Padilla. Because that tends to be the most popular 
to everybody else. Just imagine the curating opportunities that 
are possible when you get the metadata and the keywords right. 
There are some good experts that will be helpful. Dr. Hayden, 
same question to you.

[[Page 12]]

    Dr. Hayden. The Library has been really at the forefront in 
the library field with looking at how digitized material could 
be made more discoverable, because if it is not discoverable, 
it is really not accessible.
    One of the projects that I mentioned in my testimony was 
taking a 20 year project of digitizing and making available 
newspapers from all 50 states, but adding an AI instrument, a 
navigator. You can now do even more research into those 
newspapers. The photos that are in there, names, all types of 
things.
    We have seen that that has made what was just basically 
making things, putting them into digital form from the vast 
amount of analog material we have in print material going back 
hundreds of years, just basically digitizing them. That is an 
accessible aspect.
    But having tools that have people be able to get into them 
has proven very popular, especially when we allow the public to 
use material in a creative way, like our Citizen DJ program, 
where we put the material up there, music, all types of things, 
and people can use them.
    These tools can be very helpful for having people make 
those connections between our various collections--we have 
collections in 470 languages, and being able to use AI tools to 
help----
    Senator Padilla. A follow-up question to both of you. Do 
you envision institution-wide or even specific initiatives to 
always be contained within your respective institutions, or are 
there opportunities for public-private partnerships that you 
are envisioning at this point? Dr. Hayden and then Ms. Park.
    Dr. Hayden. Well, the Library of Congress and the 
Smithsonian are co-chairing a Secretariat in this coming year 
that will look at more ways that we can all collaborate and, of 
course, opportunities with the private sector when we are 
making sure there are guardrails to them.
    Senator Padilla. Correct.
    Ms. Park. I would just add that absolutely, there are 
opportunities to work with all sorts of external groups. Higher 
education, companies, all are interested in this space.
    We have formed several consortiums with different 
organizations to try to figure out if we can convene together 
to talk about these kinds of issues, to provide space for 
people to discuss the complexities of AI. Absolutely, there are 
ways for us to partner with others.
    Senator Padilla. Okay. Thank you very much.
    Chairwoman Klobuchar. Very good. Thank you. Senator 
Hagerty.
    Senator Hagerty. Thank you, Chair Klobuchar and Ranking 
Member Fischer. I appreciate you holding this hearing. Dr. 
Hayden, it is good to see you back here again--to all three of 
our witnesses. Dr. Hayden, I would like to start with you, 
though, if I might. We all know the Library of Congress houses 
the United States Copyright Office.
    Dr. Hayden. Yes.
    Senator Hagerty. I think, as a result, you are in a unique 
position to ensure the proper implementation of AI. I think, we 
all know that AI offers many promising benefits, but with its 
increasing prevalence, it is not without drawbacks.

[[Page 13]]

    I would like to take this home, and you and I have talked 
about my affinity for the songwriting and music business in my 
home State of Tennessee. We are actually home to thousands of 
performing artists there in Tennessee, songwriters, artists, 
all of whom have the potential to be greatly impacted by 
artificial intelligence.
    I am concerned that AI-generated deepfake recordings, voice 
clones, or other infringements on the works of copyright 
holders could cause serious damage to an artist's reputation 
and to their livelihood.
    Earlier this month, Governor Bill Lee announced plans for 
legislation to protect the voices of performing artists by 
adding their voices and sound to Tennessee's Protection of 
Personal Rights law.
    I am pleased that my home State of Tennessee is leading the 
Nation in providing legal protection for our artists and for 
our songwriters. My first question, Dr. Hayden, is given the 
growing pervasiveness of AI, how does the Copyright Office plan 
to protect the copyrighted works and uses of voices of 
performing artist at the federal level?
    Dr. Hayden. The Copyright Office has been in the forefront 
of looking at the issues involved with machine generated works. 
In fact, they have developed a separate web page for people who 
are creators, giving guidance on that.
    They have hosted a number of presentations and webinars, 
and our Register of Copyrights, Shira Perlmutter, has been in 
the last year in many, many international copyright sessions 
and meetings. They just issued a rule to get information, and 
they had over 10,000 responses.
    They are going to be issuing more information about what 
they are hearing from stakeholders about how that can really be 
used, but also be protecting.
    Senator Hagerty. Can I get into--let us drill into this a 
little bit more in terms of information you are getting back 
from stakeholders. In August, the Copyright Office issued a 
notice of inquiry on copyright and artificial intelligence. The 
notice sought information and comments on a number of issues, 
including the use of copyrighted works to train AI models, the 
appropriate levels of transparency and disclosure that are 
required with respect to the use of copyrighted works, the 
legal status of AI-generated outputs, and the appropriate 
treatment of AI-generated outputs that mimic the personal 
attributes of human artist.
    I am sure that many Tennessee songwriters and artists 
submitted information to the Copyright Office urging the 
implementation of several things. One, the requirements for 
obtaining appropriate licenses or authorization to ingest or 
otherwise use materials that are copyrighted, or that implicate 
a person's rights of publicity or privacy.
    Second, requiring adequate recordkeeping and auditing. 
Third, ensuring that appropriate transparency is maintained by 
AI companies. Dr. Hayden, are these remedies that I just 
listed, are those the types that the Copyright Office is 
considering?
    Dr. Hayden. Well, the office is currently reviewing those 
comments, over 10,000, and they cover the full range. What they 
are saying, and I have the information here, that they are 
really look

[[Page 14]]

ing at what could be infringement, the treatment of these types 
of AI-generated outputs, and currently they are preparing that 
report. I cannot comment further than that but they will----
    Senator Hagerty. I hope they will take into account the 
three points that I have raised that are certainly of interest 
to our industry in Tennessee. Rights to the voices and 
likenesses of performers are generally protected by state law 
rather than copyright law.
    That includes laws providing a right of publicity, privacy 
protections, and remedies for misappropriation. Given this, in 
your opinion, Dr. Hayden, are statutory changes needed to 
adequately protect the name, the image, the likeness of 
performing artists?
    Dr. Hayden. Our Register of Copyrights, Shira Perlmutter, 
is very involved with that. She will be working with Congress 
on developing any potential legislation or any other things 
that would impact the copyright law as it stands. Right now, it 
is producing works by a human being. That is the standard.
    Senator Hagerty. I certainly hope you will view my office 
as a resource and the people of Tennessee as a resource here. I 
think we can add significantly to the conversation. Thank you, 
Dr. Hayden, and I yield back my time. Thank you.
    Chairwoman Klobuchar. Very good. Thank you. Some of the 
risks of AI are already clear. I talked about what just 
happened with those robocalls and what we are seeing on--really 
in fake videos on both sides of the aisle. It is not just one 
side. I know the Republican Attorney General is investigating 
what happened in New Hampshire.
    But we know that there is going to be other security 
issues, from our infrastructure, cyber-attacks. Could you talk 
about what your agencies have done to guard against 
cybersecurity threats to ensure that your systems remain strong 
in the face of potential AI threats? Each one of you could 
answer. Dr. Hayden.
    Dr. Hayden. The Library of Congress, with the help of 
Congress, has been able to build a very, very strong 
cybersecurity foundation over a number of years.
    In fact, our CIO, Judith Conklin, is a cybersecurity 
expert. So that has been a major part of what we have been able 
to bring to looking at the risks with this particular 
technology, AI. It fits into our governance--IT governance 
framework already, and we are making sure that we are not on 
the cutting edge with AI.
    We are looking at our policies using the experiences that 
we have had. But there is a technology governance board within 
the Library, and AI now has a working group within that board 
to make sure that everything is aligned with that. We are very 
concerned about the trustworthiness of products that have AI.
    Chairwoman Klobuchar. Okay. Very good. Mr. Halpern.
    Mr. Halpern. Absolutely. Our team works very hard to make 
sure that our systems are as impregnable as possible. But we 
are also looking at some AI tools as helpful in those efforts.
    One of the things that AI excels at is pattern matching. 
They may be able to see patterns in our logs that a human being 
might not readily identify. So that is actually a benefit. One 
of the other areas that is tangentially related is in the 
privacy sphere, searching materials produced by others for 
personally identifiable information.

[[Page 15]]

    Just like many, many other organizations, we have had 
issues in the past where it has been hard for us to catch every 
incidence of PII in our materials that we make public. But 
again, these AI based tools that can recognize whether it is a 
social security number or address or another piece of PII in 
the Federal Register, the Congressional Record, or anything 
else we publish, that is an additional layer of security that 
these tools can help us with, so that we do not inadvertently 
put something out in the public domain that is not supposed to 
be there.
    Chairwoman Klobuchar. Very good. Ms. Park.
    Ms. Park. We take our cybersecurity responsibilities very 
seriously, both in terms of our own internal practices, but 
also in terms of our public domain. We do comply with NIST 
standards when it comes to our cyber profile.
    Our CIO is very much checking constantly to make sure that 
we are following those standards. Our Inspector General is also 
there to regularly check on our progress. The other thing we do 
is we regularly report and, participate in federal forums where 
you share information about the latest threats that might be 
coming through various channels, including, the emergence of AI 
and how it is being used.
    For one, for example, phishing has become much more 
sophisticated now with the emergence of AI and the ability to 
create much more easily emails and other things that appear to 
be real. There are--so we are stepping up our testing of our 
own work force in their terms of their ability to recognize 
these things. There is a lot of work going on there.
    Finally, I would mention there is an internal group led by 
our CIO right now that is really trying to develop or evolve 
our policy structure to make sure that we are prepared for 
whatever threats that might be coming from emerging 
technologies, including AI.
    Chairwoman Klobuchar. Are you all taking part in 
cybersecurity training? Do you think that will be important as 
we see this changing world, both opportunities and risks, with 
your staff?
    Dr. Hayden. Definitely----
    Chairwoman Klobuchar. Is your staff doing--okay. Yes.
    Mr. Halpern. Absolutely. We are evaluating vendors now for 
some specific AI related training so that folks understand 
these tools, and both the good and the bad that they can bring.
    Ms. Park. We have an annual training requirement for all 
staff who have access to our network.
    Chairwoman Klobuchar. Okay. Dr. Hayden, Senator Hagerty 
talked to you a bit about copyrights, voices. When you appeared 
before the Committee last year, we discussed the Copyright 
Office's new initiative to examine copyright law and policy 
issues raised by AI. What are the lessons learned by the 
Copyright Office over the past year? Have they continued to see 
increases in the amount of copyright applications for AI-
generated content?
    Dr. Hayden. Yes, they have. They have actually had a number 
of cases. There have been, as I said, about 100 works that have 
had AI-generated and human authorship, and they have been 
registered by the office.

[[Page 16]]

    They are also taking a leadership role internationally when 
looking at what is the proportion of the AI-generated content 
and human, as well. They have definitely been on the forefront 
of looking at that. But right now, they have rejected several 
applications because of the amount of AI authorship.
    Chairwoman Klobuchar. One last question of you, and then I 
will have just one more. Your written testimony highlights the 
work being done to test potential AI tools for the National 
Library Service for the Blind and Print Disabled, including 
improving the accessibility of book descriptions.
    We have talked about this way back, some of the work we 
have done with the blind. What guardrails has the Library put 
in place to ensure the content being produced by these tools is 
accurate for its users?
    Dr. Hayden. That is part of the LC Labs, and working 
directly with a service unit, NLS, National Library Service for 
the Blind and Print Disabled. They have had quite a bit of 
success with the e-reader rollout.
    They are looking at how AI can be used to help this 
population. There are some exciting opportunities, but they are 
working hand in hand to make sure that what is being presented 
to that community is trustworthy.
    You can see that that is the main theme, I think for all of 
us, that we want to use this tool for a number of reasons, for 
efficiency, effectiveness, but also we have to make sure that 
there is that oversight of use.
    Chairwoman Klobuchar. Okay. Last, Ms. Park, you talked 
about the assembling the group of experts from across the 
Smithsonian to identify best practices, opportunities. Could 
you talk about what next steps are after that, what the 
timetable is?
    Ms. Park. Yes. This group is charged with doing a number of 
things. The governance piece is an important one, reviewing 
policies, and also finding what is working within the units. We 
have multiple museums and research institutes.
    Figuring out where we can do some things to scale rather 
than have individual units try to do their own thing where we 
can, for efficiency sake. I have asked for them to do their 
initial work and come back with some recommendations for what 
we might need to do in the short term.
    What are some of the immediate stopgaps that need to be put 
in place, whether it is governance related or a best practice 
related issue that we want to share. Then I suspect there will 
be some longer term things that will be--take a little longer 
to address, whether it is systems related, business process, or 
applications. I expect it to be a bit of a phased approach.
    Chairwoman Klobuchar. My last question is, since Judy 
Garland's shoes came up, do you know where she was born, Judy 
Garland?
    Ms. Park. Oh, no, this is a quiz.
    Chairwoman Klobuchar. Yes. Well, Grand Rapids, Minnesota.
    Ms. Park. Yes, I was going to say--I was going to say 
Minnesota.
    Chairwoman Klobuchar. They just sort of solved the crime of 
the missing shoes that were there, which I would just like to 
note

[[Page 17]]

for the record, finally. That they have been returned. It is a 
sordid story. Yes, but they are back. They are back.
    In any case, thank you very much. Do you want to ask any 
other questions, Senator Fischer? We have a vote that has been 
called and thank you for the work that you are doing, 
continuing to do, leading your agencies.
    I speak for myself, I know other Members of the Committee, 
we feel good about your stewardship of these very important 
agencies, and we look forward to working with you on the 
important work you are doing.
    As we discuss opportunities, risks, I am sure we will have 
more discussions about this. The hearing record will remain 
open for one week, and we are adjourned. Thank you.
    [Whereupon, at 3:12 p.m., the hearing was adjourned.]


[[Page (18)]]



                      APPENDIX MATERIAL SUBMITTED

                              ----------                              


[GRAPHICS NOT AVAILABLE IN TIFF FORMAT]

                                    [all]