Background and Need for Legislation
When used responsibly, AI systems have the potential to
improve federal government operations, enhance national and
homeland security, and streamline service delivery.
Federal agencies have already begun to leverage AI for a
variety of use cases. For instance, a February 2020 Stanford
University report submitted to the Administrative Conference of
the United States (ACUS) found that ``nearly half of the
federal agencies studied have experimented with AI and related
machine learning tools.''\1\ The report emphasized that the
``use of AI-based tools to support government decision-making,
implementation, and interaction--what could be called
`algorithmic governance'--already spans the work of the modern
administrative state.''\2\
---------------------------------------------------------------------------
\1\David Freeman Engstrom, Daniel E. Ho, Catherine M. Sharkey, and
Mariano-Florentino Cuellar, Government by Algorithm: Artificial
Intelligence in Federal Administrative Agencies, (Feb, 2020), https://
law.stanford.edu/wp-content/uploads/2020/02/ACUS-AI-Report.pdf.
\2\Id.
---------------------------------------------------------------------------
In a more recent December 2023 report, the U.S. Government
Accountability Office (GAO) surveyed twenty-three federal
agencies to better understand their current and planned use of
AI.\3\ GAO found that twenty of the surveyed agencies were
using AI and that, collectively, these agencies reported
approximately 200 instances of AI use and approximately 1,000
instances of planned AI use.\4\ AI.gov also publicizes AI use
cases across the federal government and includes a portal for
professionals and students to join the national AI talent
surge.\5\
---------------------------------------------------------------------------
\3\U.S. Gov't accountability off., GAO-24-105980, Artificial
Intelligence: Agencies Have Begun Implementation but Need to Complete
Key Requirements, (Dec 12, 2023), https://www.gao.gov/products/gao-24-
105980.
\4\Id.
\5\AI.gov, http://ai.gov.
---------------------------------------------------------------------------
In a number of cases, federal agencies have successfully
used AI to better achieve their missions. For instance, the
Department of Homeland Security (DHS) Immigration and Customs
Enforcement uses AI to ``streamline the process of correcting
data entry errors.''\6\ The General Services Administration
(GSA) uses a Solicitation Review Tool to intake and review all
SAM.gov data to ensure that solicitations contain compliance
language, flagging those that don't as ``non-compliant''.\7\
Further, the State Department, Department of Justice, and
Centers for Disease Control and Prevention have all reportedly
used machine-learning models to search for information in
government record repositories.\8\
---------------------------------------------------------------------------
\6\Artificial Intelligence Use Case Inventory, Dept. of Homeland
Security.
\7\AI Inventory, Tech at GSA, General Services Administration.
\8\Lewis Kamb, Some U.S. government agencies are testing out AI to
help fulfill public records requests, NBC News (Aug. 1, 2023).
---------------------------------------------------------------------------
These use cases vary in both application and maturity but
nonetheless indicate that the federal government has found
relevant uses for AI systems to empower existing agency
missions and streamline programs. The potential benefits of
responsible government use of AI are enormous; however
irresponsible or improper use fosters risks to individual
privacy and the fair and equal treatment of all citizens by
their government.
As agencies use AI today and identify use cases for the
future, Congress must put in place the necessary safeguards in
H.R. 7532 to protect the public's privacy, civil rights, and
civil liberties. The public needs to know that their government
agency missions and programs have mature policies in place to
capitalize on the benefits of AI while safeguarding against
`algorithmic based' decision-making that relies solely on AI
systems without the appropriate governance and transparency
policies to ensure proper and effective use. Policymakers
should instead pursue `algorithmic informed' decision-making in
support of missions and programs through proper governance of
AI systems and policy design which accounts for the inherent
limitations of AI systems within certain use cases.
The federal government already has numerous, established
policies in place governing federal information systems, data,
cybersecurity, and procurement that should be leveraged for
future AI policy development. AI policy that does not engage
with existing information system level requirements across
these policy domains will exist outside of the current
management structure for federal information systems, leading
to confusion, inefficiency, undue administrative or industry
burden, and ultimately overlapping or competing legal or policy
requirements. For instance, legislation governing federal
agency use of AI should build upon existing areas of law
governing federal information policy and security (Chapter 35,
Title 44, U.S. Code) or the acquisition of information
technology (Chapter 113, Title 40, U.S. Code)--and where
necessary, repetitive or conflicting requirements and
definitions should be repealed or harmonized.
Transparency is an important first step to inform future
policymaking.\9\ A primary challenge in effective governance of
AI systems is forming a holistic understanding of the system's
design and provenance, the context of its use for specific
applications, and the ongoing monitoring and maintenance of its
outputs. Transparency requirements around federal agency use of
AI, such as those included in H.R. 7532, can provide important
information about government AI systems to the public, internal
government management, Congress, and stakeholders impacted by
AI informed outputs.\10\
---------------------------------------------------------------------------
\9\Alex Engler, The AI Regulatory Toolbox: How Governments Can
Discover Algorithmic Harms, Brookings (2023), https://
www.brookings.edu/articles/the-ai-regulatory-toolbox-how-governments-
can-discover-algorithmic-harms/.
\10\Id.
---------------------------------------------------------------------------
There are certain challenges, however, in establishing
transparency requirements for AI systems trained on or using
public information. Faculty from the University of Copenhagen,
in a recent research article titled, ``The Right to
Transparency in Public Governance: Freedom of Information and
the Use of Artificial Intelligence by Public Agencies,'' found
that:
[F]or AI algorithms used in public governance, there
is the additional problem that the data and knowledge
used for creating the algorithm as well as the test
data often contain personal information that cannot be
made public. Transparency then has to rely on
documentation for how the data used for the design and
test has been produced instead of access to the
underlying test data.\11\
---------------------------------------------------------------------------
\11\Henrik Palmer Olsen, Thomas Troels Hildebrandt, Cornelius
Wiesener, Matthias Smed Larsen, and Asbjrn William
Ammitzbll Flugge. 2024. The Right to Transparency in Public
Governance: Freedom of Information and the Use of Artificial
Intelligence by Public Agencies. Digit. Gov.: Res. Pract. 5, 1, Article
8 (March 2024), 15 pages. https://doi.org/10.1145/3632753.
---------------------------------------------------------------------------
The challenges in gaining total transparency into AI models
points to the need to look at AI system transparency and
governance in the context of the operational environment in
which these systems reside. In other words, maintaining sound
policies over federal IT and data supporting AI systems will
enable better governance over the AI ultimately used by
agencies. Additionally, transparency policies should be
designed to enable proper governance internal to governmental
functions (e.g., enabling oversight functions internal to the
intelligence community or Congress's oversight role over
executive branch agencies) while also pursuing public
transparency requirements in order to enable the appropriate
levels of operational management, governance, and oversight
over an AI system's entire lifecycle. Such policies should be
applied in context of the complexity and corresponding risk
profile of the AI systems (i.e., simple process automation and
workflow tools should not require the same level of system
governance controls as large language models deployed with
access to sensitive public sector data sets) so as to not
unnecessarily impede agency adoption.
In recent years, Congress has passed an assortment of AI-
related legislation, and the Trump and Biden administrations
have issued a variety of executive orders and policies
regarding agency use of AI. These initiatives, which are
outlined below, have produced positive, but at times disjointed
or contradictory directives for the management of federal
agency AI systems as such policies and laws have spanned
multiple Administrations and Congresses while AI technology has
rapidly evolved.
In February 2019, President Trump issued Executive Order
13859 on Maintaining American Leadership in Artificial
Intelligence, which established the American AI Initiative
coordinated through the National Science and Technology
Council.\12\ This initiative directed implementing agencies to
consider AI as an agency research and development (R&D)
priority, to increase the non-Federal research community's
access to data and computing resources for AI R&D, and to
invest in AI-related educational and workforce development.\13\
EO 13859 also directed the OMB Director to issue guidance to
inform agency approaches to AI applications and
regulations.\14\
---------------------------------------------------------------------------
\12\Exec. Order No. 13859, 84 FR 3967 (Feb. 14, 2019).
\13\Id.
\14\Id.
---------------------------------------------------------------------------
In December 2020, President Trump issued Executive Order
13960 on Promoting the Use of Trustworthy Artificial
Intelligence in the Federal Government, which directs agencies
to adhere to a common set of nine principles to guide agency
design, development, acquisition and use of AI.\15\ EO 13960
also directs agencies to prepare, and in most cases publish,
inventories of current and planned uses of AI to foster
transparency into the federal government's use of AI.\16\
---------------------------------------------------------------------------
\15\Exec. Order No. 13960, 85 FR 78939 (Dec. 8, 2020).
\16\Id.
---------------------------------------------------------------------------
Also in December 2020, Congress enacted the AI in
Government Act of 2020 (P.L. 116-260 Division U, Title I; 40
U.S.C. 11301 note), which requires OMB to issue government-wide
guidance on agency use of AI and agency AI governance
plans.\17\ This legislation also established an AI occupational
series for the federal workforce.\18\
---------------------------------------------------------------------------
\17\Consolidated Appropriations Act, 2021, Pub. L. No. 116-260,
Div. U, Title I, AI in Govt. Act (2020).
\18\Id.
---------------------------------------------------------------------------
In January 2021, Congress enacted the National AI
Initiative Act of 2020, (P.L. 116-283), which established the
National Artificial Intelligence Initiative and required the
White House Office of Science and Technology Policy (OSTP) to
establish or designate the National Artificial Intelligence
Initiative Office to coordinate ongoing AI research,
development, and demonstration activities across the federal
government and carry out additional responsibilities with
respect to this initiative.\19\ This legislation also requires
the National Institute of Standards and Technology (NIST) to
develop voluntary standards for artificial intelligence
systems.\20\
---------------------------------------------------------------------------
\19\William M. (Mac) Thornberry National Defense Authorization Act
for Fiscal Year 2021, Pub. L. No. 116-283, Title XLVII, Div. E,
National Artificial Intelligence Initiative Act of 2020 (2021).
\20\Id.
---------------------------------------------------------------------------
In October 2022, Congress enacted the AI Training for the
Acquisition Workforce Act, (P.L. 117-40; 41 U.S.C. 1703 note),
which requires OMB and GSA to provide a training program on AI
for the program management and acquisition workforce to ensure
they have knowledge of the capabilities and risks associated
with AI.\21\
---------------------------------------------------------------------------
\21\Artificial Intelligence Training for the Acquisition Workforce
Act, Pub. L. No. 117-207 (2022).
---------------------------------------------------------------------------
In October 2022, President Biden's OSTP issued a Blueprint
for an AI Bill of Rights, which is a white paper providing non-
binding, directional guidance regarding the development and use
of AI systems that ``have the potential to meaningfully impact
the American public's rights, opportunities, or access to
critical resources or services.''\22\ The Blueprint does not
constitute U.S. government policy.\23\
---------------------------------------------------------------------------
\22\Blueprint for an AI Bill of Rights: Making Automated Systems
Work for the American People, WhiteHouse.Gov, (October 2022).
\23\Id.
---------------------------------------------------------------------------
In December 2022, Congress enacted the Advancing American
AI Act, (P.L. 117-263, Title LXXII, Subtitle B; 40 U.S.C. 11301
note), which expands the AI in Government Act of 2020 by
directing OMB to also consider Executive Order 13960, cross-
government input, and the recommendations form the National
Security Commission on Artificial Intelligence while developing
the guidance required by that legislation.\24\ This legislation
also encourages OMB to designate a centralized, public
inventory of AI use cases and requires OMB to identify four new
use cases where AI can support interagency modernization
initiatives to improve operations across organizational
boundaries, including to ``drive agency productivity
efficiencies'' or ``accelerate agency investment return''.\25\
---------------------------------------------------------------------------
\24\James M. Inhofe National Defense Authorization Act for Fiscal
Year 2023, Pub. L. No. 117-263, Title LXXII, Subtitle B, Advancing
American AI Act (2022).
\25\Id.
---------------------------------------------------------------------------
In January 2023, NIST issued an AI Risk Management
Framework (AI RMF 1.0), which is a voluntary framework
developed by NIST to inform responsible development and use of
AI systems.\26\
---------------------------------------------------------------------------
\26\National Institute of Standards and Technology, NIST AI 100-1,
Artificial Intelligence Risk Management Framework (AI RMF 1.0) (January
2023).
---------------------------------------------------------------------------
In October 2023, President Biden issued Executive Order
14110 on the Safe, Secure, and Trustworthy Development and Use
of Artificial Intelligence, which establishes eight guiding
principles and includes dozens of new reforms, guidelines,
rules, and programs across eight key workstreams. EO 14110
tasks over 50 federal entities with approximately 150 distinct
requirements.\27\ EO 14110 also includes several requirements
and guidelines for federal agency use of AI, including a
requirement that the Technology Modernization Board consider
prioritizing AI projects when awarding funds from the
Technology Modernization Fund; a requirement that agencies
implement or increase the availability of AI training programs
from employees, managers, and leadership in technology and
relevant policy fields; a requirement that GSA facilitate
access to federal government-wide acquisition solutions for AI
services and products, including generative AI and computing
infrastructure; and a requirement that the White House Deputy
Chief of Staff for Policy convene an AI and Technology Talent
Task Force to accelerate and monitor the hiring and training of
AI talent in the federal government.
---------------------------------------------------------------------------
\27\Rishi Bommasani et al., Decoding the White House AI Executive
Order's Achievements, Stanford University Human-Centered Artificial
Intelligence (Nov. 2, 2023), available at: https://hai.stanford.edu/
news/decoding-white-house-ai-executive-orders-achievements.
---------------------------------------------------------------------------
In March 2024, OMB issued M-24-10, a memorandum for the
heads of executive departments and agencies on Advancing
Governance, Innovation, and Risk Management for Agency Use of
Artificial Intelligence, which directs agencies to advance AI
governance and innovation while managing risks from the use of
AI in the Federal Government.\28\
---------------------------------------------------------------------------
\28\Office of Management and Budget, M-24-10, Advancing Governance,
Innovation, and Risk Management for Agency Use of Artificial
Intelligence, (March 2024).
---------------------------------------------------------------------------
In September 2024, OMB issued M-24-18, a memorandum for the
heads of executive departments and agencies on Advancing the
Responsible Acquisition of Artificial Intelligence in
Government, which directs agencies to improve their capacity
for the responsible acquisition of AI.\29\
---------------------------------------------------------------------------
\29\Office of Management and Budget, M-24-18, Advancing the
Responsible Acquisition of Artificial Intelligence in Government,
(September 2024).
---------------------------------------------------------------------------
While these initiatives have produced positive directives
for federal agencies, important steps remain to achieve a more
useful and comprehensive governance framework for federal
agency AI use. For example, the Center for Democracy and
Technology stated in a recent review of federal agency AI
inventories that ``[t]he information provided by each agency is
inconsistent and unclear, making it difficult for the public to
understand exactly how the use of AI impacts them.''\30\
---------------------------------------------------------------------------
\30\Bowman Cooper, Like Looking for a Needle in an AI Stack, Center
for Democracy and Technology (2023). https://cdt.org/insights/like-
looking-for-a-needle-in-an-ai-stack/.
---------------------------------------------------------------------------
H.R. 7532 builds on the core themes of these past efforts
by centrally codifying federal agency governance and
responsible use policies while consolidating and streamlining
other existing federal agency AI laws in order to reduce
duplicative or contradictory directives to federal agencies,
simplify compliance, and provide a strong foundation for future
legislative efforts. In doing so, the bill focuses government
resources
Section-by-Section Analysis
Section 1. Short title
Section 2. Establishment of Federal agency artificial intelligence
system governance requirements
Subsection (a) adds a new Subchapter IV in Chapter 35,
Title 44 of the U.S. Code to establish federal agency AI system
governance requirements, including the following new sections:
``Sec. 3591. Purposes. Codifies nine
directives for agency heads to adhere to when
designing, developing, acquiring, using, managing, or
conducting oversight over AI in the Federal Government.
``Sec. 3592. Definitions. Applies existing
definitions under 44 U.S.C. 3502 and defines additional
terms including: ``artificial intelligence,''
``artificial intelligence system,'' ``federal
artificial intelligence system,'' ``federal information
system,'' and ``national security system.''
``Sec. 3593. Authority and functions of the
Director. Places the Director of the Office of
Management and Budget (OMB) in charge of overseeing
government-wide use of federal AI systems, including
implementation of the purposes outlined in section
3591, as well as:
Developing, coordinating, and
overseeing the implementation of policies,
principles, standards, and guidelines to ensure
appropriate use of federal AI systems for the
protection of civil rights and civil liberties,
and conformity with existing Privacy Act (5
U.S.C. 552a) and federal information system
protections.
D Note: Recodifies existing
provisions of section 7224(d)(1)(B) of
the Advancing American AI Act (P.L.
117-263, Division G, Title LXXII,
Subtitle B; 40 U.S.C. 11301 note).
Requiring agencies to identify
and provide protections and safeguards for the
use of federal AI systems commensurate with the
risk in a consistent manner with the standards
promulgated under the National Institute of
Standards and Technology Act (15 U.S.C. 278h-
1).
Recommending approaches to
removing barriers to agency use of AI
technologies while protecting privacy, civil
liberties, civil rights, and economic and
national security, as well as, identifying,
assessing, and mitigating any discriminatory
impact or bias.
D Note: Recodifies existing
provisions of section 104(a)(1)-(3) of
the AI in Government Act (P.L. 116-260,
Division U, Title I; 40 U.S.C. 11301
note) and section 7224(d)(1) of the
Advancing American AI Act (P.L. 117-
263, Division G, Title LXXII, Subtitle
B; 40 U.S.C. 11301 note).
Providing agencies with guidance
to establish a plain language notification and
appeals process that conforms with existing
statutory requirements regarding the protection
of agency records for individuals or entities
impacted by an agency determination based on a
federal AI system output.
Providing guidance and a
template for the agency AI plans described in
section 3594 and issuing guidance to help
agencies establish the AI governance charters,
including defining high-risk Federal artificial
intelligence systems, described in section
3595.
``Sec. 3594. Federal agency
responsibilities. Requires the head of each agency to
ensure that federal AI system management processes are
integrated with agency planning and that senior agency
officials implement necessary policies and procedures,
with the agency Chief Information Officer (CIO), in
coordination with other appropriate senior agency AI
officials, being the responsible agency official for
compliance with this subchapter.
Requires a public plan providing
information about the agency's federal AI
system governance policies and procedures,
including the inventory of AI use cases as
required by section 7225(a) of the Advancing
American AI Act (P.L. 117-263; 40 U.S.C. 11301
note).
Requires the modification of an
agency's appeals process for individuals or
entities impacted by federal agency
determinations that were substantively and
meaningfully augmented by a federal AI system.
Requires, in accordance with OMB
guidance and section 3595, the implementation
of regularly updated AI governance charters
that are submitted to the Federal Register and
the General Services Administration (GSA) for
publication under section 3596.
Requires regular training
programs for relevant agency employees on
federal AI system management requirements,
which may be integrated into training
requirements established by the A.I. Training
for the Acquisition Workforce Act (P.L. 117-
207).
``Sec. 3595. Agency AI Governance Charters.
Requires agency heads to ensure that an accurate and
complete AI governance charter is established for each
federal AI system in use by an agency that has been
designated as high-risk (as defined by OMB guidance) or
interacts with a record maintained on an individual (as
defined under the Privacy Act; 5 U.S.C. 552a)--with a
required annual review and requirement that a new
charter be issued within 30 days of any major system
changes--and specifies the required AI governance
charter contents, including:
The bureau, department, or
office using or operating the federal AI
system, and the contact information for the
designated agency official responsible for
maintenance of the system.
Information about how the
federal AI system was developed and funded,
including the training, validation, and testing
of the system.
Information about ongoing
oversight and maintenance of the system,
including the cadence of system testing,
validation, monitoring, and auditing, and
performance metrics considered by the agency.
Information about how the system
is used, the data or data assets produced by
the system, and whether the system was trained
on, uses, or produces a federal record
maintained on an individual.
Subsection (d) requires the AI
governance charter of each agency to be made
public on the agency webpage and noticed in the
Federal Register but allows an agency head to
waive the publication requirement with approval
by the OMB Director and notification to
Congress and the agency Inspector General, or
to protect properly classified national
security information.
Subsection (e) exempts federal
AI systems used solely for research and
development or systems used in national
security systems from the requirements of this
section.
Sec. 3596. AI Governance Charter Inventory.
This section requires the GSA Administrator to maintain
a single, public, online interface for centrally
cataloging agency AI governance charters in a machine-
readable and open format for bulk download, which shall
be known as the ``Federal AI System Inventory.''
``Sec. 3597. Independent evaluation.
Requires IGs to evaluate the federal AI governance
policies and practices of the agency every two years
and submit to the agency head, the OMB Director, and
Congress, a report which may include a classified annex
and directs the Comptroller General to periodically
evaluate and submit to Congress a report on the
effectiveness of agency federal AI system governance
policies and practices.
Subsection (b) requires the OMB Director--in consultation
with NIST, GSA, and the Office of Science and Technology Policy
(OSTP)--to issue a memo to agencies one year after enactment
(to be updated as necessary every two years and then
periodically after 10 years) establishing guidelines for
implementing the requirements of this Act's newly established
Subchapter IV.
Subsection (c) updates the Privacy Act (5 U.S.C. 552a) to
include listing of any related AI governance charters in an
agency system of records notice and to ensure the security,
confidentiality, and integrity of federal AI system records.
Subsection (d) repeals provisions of the AI in Government
Act and the Advancing American AI Act which have been
recodified into the new Subchapter IV established by subsection
(a) of this Act.
Subsection (e) requires that within six months after the
issuance of OMB guidance under subsection (b) the Federal
Acquisition Regulation (FAR) will be updated to reflect the
amendments made in this section and require contractors and
subcontractors that work on federal AI systems to provide
agencies with the information necessary for compliance with
this Act.
Subsection (f) restricts the ability of agencies to take
actions not authorized by this Act, specifies protections of
individuals' constitutional and privacy rights, and clarifies
that nothing in the Act should be construed as requiring the
public disclosure of information that could otherwise be
withheld under the Freedom of Information Act (5 U.S.C.
522(b)).
Subsection (g) clarifies that the definition of ``agency''
and ``Director''.
Legislative History
H.R. 7532, the Federal AI Governance and Transparency Act,
was introduced on March 5, 2024, by Representative James Comer.
The following Representatives are cosponsors of the bill: Jamie
Raskin (D-MD), Nancy Mace (R-SC), Alexandria Ocasio-Cortez (D-
NY), Clay Higgins (R-LA), Gerald E. Connolly (D-VA), Nicholas
A. Langworthy (R-NY), and Ro Khanna (D-CA). The bill was
referred to the Committee on Oversight and Accountability. The
Committee on Oversight and Accountability held a hearing
related to and used for development and consideration of the
bill on September 14, 2023. The Committee considered H.R. 7532
at a business meeting on March 7, 2024, and ordered the bill as
amended favorably reported by a recorded vote.
Committee Consideration
On March 7, 2024, the Committee met in open session and
ordered the bill, H.R. 7532, favorably reported with an
amendment in the nature of a substitute, by a roll call vote of
36-3, a quorum being present.
Roll Call Votes
In compliance with clause 3(b) of rule XIII of the Rules of
the House of Representatives, the following roll call vote
occurred during the Committee's consideration of H.R. 7532:
The first and only roll call vote was on final passage of
H.R. 7532. The bill was agreed to in a recorded vote of 36-3.
[GRAPHIC(S) NOT AVAILABLE IN TIFF FORMAT]
Explanation of Amendments
During Committee consideration of the bill, Representative
James Comer (R-KY), Chairman of the Committee, offered an
amendment in the nature of a substitute that would make certain
technical changes to the bill. The amendment in the nature of a
substitute passed by voice vote.
List of Related Committee Hearings
In accordance with clause 3(c)(6) of rule XIII of the Rules
of the House of Representatives, (1) The following hearing was
used to develop or consider H.R. 7532:
On September 14, 2023, the Subcommittee on Cybersecurity,
Information Technology, and Government Innovation held a
hearing titled ``How are Federal Agencies Harnessing Artificial
Intelligence?'' with Dr. Arati Prabhakar, Director, White House
Office of Science and Technology Policy; Dr. Craig Martell,
Chief Digital and AI Officer, Department of Defense; and Mr.
Eric Hysen, Chief Information Officer, Department of Homeland
Security.
(2) The following related hearings were held:
On March 8, 2023, the Subcommittee on Cybersecurity,
Information Technology, and Government Innovation held a
hearing titled ``Advances in AI: Are We Ready for a Tech
Revolution?'' with Dr. Eric Schmidt, Chair, Special Competitive
Studies Project; Dr. Aleksander Madry, Director, MIT Center for
Deployable Machine Learning, and Cadence Design Systems
Professor of Computing, Massachusetts Institute of Technology;
Dr. Scott Crowder, Vice President, IBM Quantum, and CTO, IBM
Systems, Technical Strategy and Transformation IBM; and Ms.
Merve Hickok, Senior Research Director, Center for AI and
Digital Policy.
On June 22, 2023, the Subcommittee on Cybersecurity,
Information Technology, and Government Innovation held a
hearing titled ``Using Cutting-Edge Technologies to Keep
America Safe'' with Mr. Ryan Rawding, Vice President of
Business Development, Pangiam; Mr. Wahid Nawabi, Chairman,
President, and Chief Executive Officer, AeroVironment; Mr.
Gavin Kenneally, Chief Executive Officer, Ghost Robotics; and
Dr. Benjamin Boudreaux, Policy Researcher and Professor of
Policy Analysis, Pardee RAND Graduate School, RAND Corporation.
On December 6, 2023, the Subcommittee on Cybersecurity,
Information Technology, and Government Innovation held a
hearing titled ``White House Policy on AI'' with Mr. Ross
Nodurft, Executive Director, Alliance for Digital Innovation;
Mr. Samuel Hammond, Senior Economist, Foundation for American
Innovation; Ms. Kate Goodloe, Managing Director BSA | The
Software Alliance; Dr. Daniel Ho, William Benjamin Scott and
Luna M. Scott Professor of Law, Senior Fellow, Stanford
Institute for Human-Centered AI, Stanford Law School; and Dr.
Rumman Chowdhury, Responsible AI Fellow, Berkman Klein Center
for Internet & Society, Harvard University.
On January 17, 2024, the Subcommittee on Cybersecurity,
Information Technology, and Government Innovation held a
hearing titled ``Toward an AI Ready Workforce'' with Dr.
William Scherlis, Professor of Computer Science, Carnegie
Mellon University; Ms. Timi Hadra, Client Partner and Senior
Executive for West Virginia, IBM; Dr. Richard Levin, Former
President, Yale University and Senior Advisor, Coursera; and
Dr. Costis Toregas, Director, Cyber Security and Privacy
Research Institute, The George Washington University.
On March 21, 2024, the Subcommittee on Cybersecurity,
Information Technology, and Government Innovation held a
hearing titled ``White House Overreach on AI'' with Ms.
Jennifer Huddleston, Technology Policy Research Fellow, Cato
Institute; Mr. Adam Thierer, Resident Senior Fellow, Technology
& Innovation, R Street Institute; Mr. Neil Chilson, Head of AI
Policy The Abundance Institute; and Dr. Nicol Turner Lee,
Senior Fellow, Governance Studies and Director, Center for
Technology Innovation, Brookings Institution.
Statement of Oversight Findings and Recommendations of
the Committee
In compliance with clause 3(c)(1) of rule XIII and clause
(2)(b)(1) of rule X of the Rules of the House of
Representatives, the Committee's oversight findings and
recommendations are reflected in the Background and Need for
Legislation section above.
Statement of General Performance Goals and Objectives
In accordance with clause 3(c)(4) of rule XIII of the Rules
of the House of Representatives, the Committee's performance
goals or objectives of this bill are to amend chapter 35 of
title 44, United States Code, to establish Federal AI system
governance requirements, and for other purposes.
Application of Law to the Legislative Branch
Section 102(b)(3) of Public Law 104-1 requires a
description of the application of this bill to the legislative
branch where the bill relates to the terms and conditions of
employment or access to public services and accommodations.
This bill does not relate to employment or access to public
services and accommodations in the legislative branch.
Duplication of Federal Programs
In accordance with clause 3(c)(5) of rule XIII of the Rules
of the House of Representatives, no provision of this bill
establishes or reauthorizes a program of the Federal Government
known to be duplicative of another Federal program, a program
that was included in any report from the Government
Accountability Office to Congress pursuant to section 21 of
Public Law 111-139, or a program related to a program
identified in the most recent Catalog of Federal Domestic
Assistance.
Federal Advisory Committee Act Statement
The Committee finds that this legislation does not direct
the establishment of advisory committees within the definition
of Section 5(b) of the appendix to title 5, U.S.C.
Unfunded Mandates Reform Act Statement
Pursuant to section 423 of the Congressional Budget Act of
1974 the Committee has included a letter received from the
Congressional Budget Office below.
Earmark Identification
This bill does not include any congressional earmarks,
limited tax benefits, or limited tariff benefits as defined in
clause 9 of rule XXI of the Rules of the House of
Representatives.
Committee Cost Estimate
Pursuant to clause 3(d) of rule XIII of the Rules of the
House of Representatives, the Committee includes below a cost
estimate of the bill prepared by the Director of the
Congressional Budget Office under section 402 of the
Congressional Budget Act of 1974.
New Budget Authority and Congressional Budget Office
Cost Estimate
Pursuant to clause 3(c)(2) of rule XIII of the Rules of the
House of Representatives and section 308(a) of the
Congressional Budget Act of 1974, and pursuant to clause
3(c)(3) of rule XIII of the Rules of the House of
Representatives, the cost estimate prepared by the
Congressional Budget Office and submitted pursuant to section
402 of the Congressional Budget Act of 1974 is as follows:
[GRAPHIC(S) NOT AVAILABLE IN TIFF FORMAT]
H.R. 7532 would require federal agencies to develop and
publish charters governing the operations, maintenance, and
transparency of their artificial intelligence (AI) programs.
The bill would codify the responsibilities of the Office of
Management and Budget (OMB) in producing government-wide
policies for the safe and secure adoption of AI technology. The
bill also would require agency inspectors general and the
Government Accountability Office to report to the Congress
every two years after enactment on the effectiveness of those
efforts.
The Administration has issued executive orders and
memoranda concerning the creation of federal AI programs. CBO
expects those actions will satisfy most of the requirements of
the bill. Using information on the costs of similar plans and
reports, CBO estimates that satisfying the governance charter
and reporting requirements would cost $6 million over the 2024-
2029 period. Such spending would be subject to the availability
of appropriated funds.
Enacting H.R. 7532 could affect direct spending by some
agencies that are allowed to use fees, receipts from the sale
of goods, and other collections to cover operating costs. CBO
estimates that any net changes in direct spending by those
agencies would be negligible because most of them can adjust
amounts collected to reflect changes in operating costs.
The costs of the legislation, detailed in Table 1, fall
within budget function 800 (general government).
TABLE 1.--ESTIMATED INCREASES IN SPENDING SUBJECT TO APPROPRIATION UNDER H.R. 7532
----------------------------------------------------------------------------------------------------------------
By fiscal year, millions of dollars--
----------------------------------------------------------------
2024 2025 2026 2027 2028 2029 2024-2029
----------------------------------------------------------------------------------------------------------------
Estimated Authorization........................ * 4 * 1 * 1 6
Estimated Outlays.............................. * 4 * 1 * 1 6
----------------------------------------------------------------------------------------------------------------
* = between zero and $500,000.
The CBO staff contact for this estimate is Aldo Prosperi.
The estimate was reviewed by Christina Hawley Anthony, Deputy
Director of Budget Analysis.
Phillip L. Swagel,
Director, Congressional Budget Office.
Changes in Existing Law Made by the Bill, as Reported
In compliance with clause 3(e) of rule XIII of the Rules of
the House of Representatives, changes in existing law made by
the bill, as reported, are shown as follows (existing law
proposed to be omitted is enclosed in black brackets, new
matter is printed in italics, and existing law in which no
change is proposed is shown in roman):
Changes in Existing Law Made by the Bill, as Reported
In compliance with clause 3(e) of rule XIII of the Rules of
the House of Representatives, changes in existing law made by
the bill, as reported, are shown as follows (existing law
proposed to be omitted is enclosed in black brackets, new
matter is printed in italics, and existing law in which no
change is proposed is shown in roman):
TITLE 44, UNITED STATES CODE
PART A--GENERAL
* * * * * * *
CHAPTER 35--COORDINATION OF FEDERAL
INFORMATION POLICY
SUBCHAPTER I--FEDERAL INFORMATION POLICY
Sec.
* * * * * * *
Subchapter IV--Artificial Intelligence System Governance
3591. Purposes.
3592. Definitions.
3593. Authority and functions of the Director.
3594. Federal agency responsibilities.
3595. Agency AI Governance Charters.
3596. AI Governance Charter Inventory.
3597. Independent evaluation.
* * * * * * *
SUBCHAPTER IV--ARTIFICIAL INTELLIGENCE SYSTEM
GOVERNANCE
Sec. 3591. Purposes
The purposes of this subchapter, with respect to the design,
development, acquisition, use, management, and oversight of
artificial intelligence in the Federal Government, are to
ensure the following:
(1) Actions that are consistent with the Constitution
and any other applicable law and policy, including
those addressing freedom of speech, privacy, civil
rights, civil liberties, and an open and transparent
Government.
(2) Any such action is purposeful and performance-
driven, including ensuring the following:
(A) Such action promotes the consistent and
systemic treatment of all individuals in a
fair, just, and impartial manner.
(B) The public benefits of such action
significantly outweigh the risks.
(C) The risks and operations of such action
do not unfairly and disproportionately benefit
or harm an individual or subgroup of the
public.
(D) The risk of such action is assessed and
responsibly managed, including before the use
of artificial intelligence.
(3) Any application of artificial intelligence is
consistent with the use cases for which the artificial
intelligence was trained, and the deployers of such
application promote verifiably accurate, ethical,
reliable, and effective use.
(4) The safety, security, and resiliency of
artificial intelligence applications, including
resilience when confronted with any systematic
vulnerability, adversarial manipulation, and other
malicious exploitation.
(5) The purpose, operations, risks, and outcomes of
artificial intelligence applications are sufficiently
explainable and understandable, to the extent
practicable, by subject matter experts, users, impacted
parties, and others, as appropriate.
(6) Such action is responsible and accountable,
including by ensuring the following:
(A) Human roles and responsibilities are
clearly defined, understood, and appropriately
assigned.
(B) Artificial intelligence is used in a
manner consistent with the purposes described
in this section and the purposes for which each
use of artificial intelligence is intended.
(C) Such action, as well as relevant inputs
and outputs of artificial intelligence
applications, are well documented and
accountable.
(7) Responsible management and oversight by ensuring
the following:
(A) Artificial intelligence applications are
regularly tested against the purposes described
in this section.
(B) Mechanisms are maintained to supersede,
disengage, or deactivate applications of
artificial intelligence that demonstrate
performance or outcomes that are inconsistent
with the intended use or this subchapter.
(C) Engagement with impacted communities.
(8) Transparency in publicly disclosing relevant
information regarding the use of artificial
intelligence to appropriate stakeholders, to the extent
practicable and in accordance with any applicable law
and policy, including with respect to the protection of
privacy, civil liberties, and of sensitive law
enforcement, national security, trade secrets or
proprietary information, and other protected
information.
(9) Accountability for the following:
(A) Implementing and enforcing appropriate
safeguards necessary to comply with the
purposes described in this section and the
requirements of this subchapter, for the proper
use and functioning of the applications of
artificial intelligence.
(B) Monitoring, auditing, and documenting
compliance with those safeguards, as
appropriate.
(C) Providing appropriate training to all
agency personnel responsible for the design,
development, acquisition, use, management, and
oversight of artificial intelligence.
Sec. 3592. Definitions
In this subchapter:
(1) In general.--Except as provided in paragraph (2),
the definitions under sections 3502 shall apply to this
subchapter.
(2) Additional definitions.--In this subchapter:
(A) Administrator.--The term
``Administrator'' means the Administrator of
General Services.
(B) Appropriate congressional committees.--
The term ``appropriate congressional
committees'' means the Committee on Oversight
and Accountability of the House of
Representatives and the Committee on Homeland
Security and Governmental Affairs of the
Senate.
(C) Artificial intelligence.--The term
``artificial intelligence'' has the meaning
given the term in section 238(g) of the John S.
McCain National Defense Authorization Act for
Fiscal Year 2019 (Public Law 115-232; 10 U.S.C.
note prec. 4061).
(D) Artificial intelligence system.--The term
``artificial intelligence system'' means any
data system, software, application, tool, or
utility that operates in whole or in part using
dynamic or static machine learning algorithms
or other forms of artificial intelligence,
whether--
(i) the data system, software,
application, tool, or utility is
established primarily for the purpose
of researching, developing, or
implementing artificial intelligence
technology; or
(ii) artificial intelligence
capability is integrated into another
system or business process, operational
activity, or technology system.
(E) Federal artificial intelligence system.--
The term ``Federal artificial intelligence
system'' means an artificial intelligence
system used in connection with a Federal
information system.
(F) Federal information system.--The term
``Federal information system'' has the meaning
given the term in section 11331(g) of title 40.
(G) National security system.--The term
``national security system'' has the meaning
given that term in section 3552(b) of title 44.
Sec. 3593. Authority and functions of the Director
The Director shall oversee the design, development,
acquisition, use, management, and oversight of Federal
artificial intelligence systems by agencies to implement the
purposes described in section 3591. In performing such
oversight, the Director shall do the following:
(1) Develop, coordinate, and oversee the
implementation of policies, purposes, standards, and
guidelines to ensure appropriate use of Federal
artificial intelligence systems and the protection of
civil rights, civil liberties, and privacy, including
in conformity with section 552a of title 5 and other
applicable laws, as well as the integrity of Federal
information systems and information technology in
accordance with the other requirements of this chapter.
(2) Oversee agency compliance with the requirements
of this subchapter, including through any authorized
enforcement action under section 11303(b)(5) of title
40 to ensure agency accountability and compliance.
(3) Issue and update, as necessary, guidance to
agencies to take steps to advance the governance of
Federal artificial intelligence systems, manage risk,
and remove relevant barriers to innovation, consistent
with the requirements of this subchapter and, as
appropriate the standards promulgated under section 22A
of the National Institute of Standards and Technology
Act (15 U.S.C. 278h-1) pursuant to section 5302 of the
William M. (Mac) Thornberry National Defense
Authorization Act for Fiscal Year 2021 (15 U.S.C. 9441)
that addresses the following:
(A) The development of policies regarding
Federal acquisition, procurement, and use by
agencies regarding artificial intelligence,
including an identification of the
responsibilities of agency officials managing
the use of such technology.
(B) The ownership and protection of data and
other information created, used, processed,
stored, maintained, disseminated, disclosed, or
disposed of by a contractor or subcontractor
(at any tier) on behalf of the Federal
Government.
(C) The protection of training data,
algorithms, and other components of any Federal
artificial intelligence system against misuse,
unauthorized alteration, degradation, or being
rendered inoperable.
(D) The removal of barriers to responsible
agency use of artificial intelligence, such as
information technology, data, workforce, and
budgetary barriers, in order to promote the
innovative application of those technologies
while protecting privacy, civil liberties,
civil rights, and economic and national
security.
(E) The establishment of best practices for
identifying, assessing, and mitigating any
discrimination in violation of title VI of the
Civil Rights Act of 1964 (42. U.S.C. 2000d et
seq.), or any unintended consequence of the use
of artificial intelligence, including policies
to--
(i) identify data used to train
artificial intelligence;
(ii) identify data analyzed or
ingested by Federal artificial
intelligence systems used by the
agencies; and
(iii) require periodic evaluation of
Federal artificial intelligence
systems, as appropriate.
(4) Issue guidance for agencies to establish a plain
language notification process, as necessary and
appropriate and in conformity with applicable law,
including section 552a of title 5, for individuals or
entities impacted by an agency determination that has
been based solely on an output from, or substantively
and meaningfully informed, augmented, or assisted by a
Federal artificial intelligence system, including the
contents of any notice, including examples of what the
notice may look like in practice.
(5) Issue guidance for agencies to review their
appeals process and to make modifications, as necessary
and appropriate, to account for determinations made
solely by or substantively and meaningfully informed,
augmented, or assisted by a Federal artificial
intelligence system, including guidance on how an
agency provides the impacted individual or entity the
opportunity for an alternative review independent of
the Federal artificial intelligence system, as
appropriate.
(6) Provide guidance and a template for the required
contents of the agency plans described in section
3594(6) that uses a uniform resource locator that is in
a consistent format across agencies such as the format
``agencyname.gov/AI''.
(7) Issue guidance, including a uniform required
submission format and criteria for updating entries
after significant changes, for the establishment of
agency AI governance charters under section 3595,
including defining high-risk Federal artificial
intelligence systems, and publication under section
3596.
Sec. 3594. Federal agency responsibilities
The head of each agency shall do the following:
(1) Comply with the requirements of this subchapter
and related policies, purposes, standards, and
guidelines, including those under section 552a of title
5 and in guidance issued by the Director under section
3593.
(2) Ensure that Federal artificial intelligence
system management processes are integrated with agency
strategic, operational, data, workforce planning, and
budgetary planning processes, and other requirements
under this chapter.
(3) Ensure that senior agency officials, including
the Chief Information Officer, the Chief Data Officer,
and the senior agency official for privacy, implement
policies and procedures regarding Federal artificial
intelligence systems under the control of such
officers, assess and reduce any risks to such systems
to an acceptable level, and periodically assess and
validate management procedures and controls to ensure
effective implementation of this subchapter.
(4) Delegate to the agency Chief Information Officer
established under section 3506 (or comparable official
in an agency not covered by such section) the primary
authority and accountability to ensure compliance with
the agency requirements under this subchapter in
coordination with any other appropriate senior agency
official designated by the head of the agency.
(5) Ensure that contracts for the acquisition and
procurement of a Federal artificial inte