Minority Views
[To accompany H.R. 4801]
The Committee on Financial Services, to whom was referred
the bill (H.R. 4801) to establish AI Innovation Labs that
permit certain persons to experiment with artificial
intelligence without expectation of enforcement actions, having
considered the same, reports favorably thereon with an
amendment and recommends that the bill as amended do pass.
CONTENTS
Page
Purpose and Summary.............................................. 7
Background and Need for Legislation.............................. 7
Committee Consideration.......................................... 8
Related Hearings................................................. 9
Committee Votes.................................................. 10
Committee Oversight Findings..................................... 16
Performance Goals and Objectives................................. 16
Committee Cost Estimate.......................................... 16
New Budget Authority and CBO Cost Estimate....................... 16
Unfunded Mandates Statement...................................... 16
Earmark Statement................................................ 16
Federal Advisory Committee Act Statement......................... 17
Applicability to the Legislative Branch.......................... 17
Duplication of Federal Programs.................................. 17
Section-by-Section Analysis of the Legislation................... 17
Changes in Existing Law Made by the Bill, as Reported............ 18
Documents Included by Unanimous Consent.......................... 19
Minority Views................................................... 21
The amendment is as follows:
Strike all after the enacting clause and insert the
following:
SECTION 1. SHORT TITLE.
This Act may be cited as the ``Unleashing AI Innovation in Financial
Services Act''.
SEC. 2. DEFINITIONS.
In this section:
(1) AI test project.--The term ``AI test project'' means a
financial product, service, or activity--
(A) that falls under the jurisdiction of a financial
regulatory agency;
(B) that makes substantial use of artificial
intelligence;
(C) that is, or may be, subject to a Federal
regulation or Federal statute; and
(D) for which a regulated entity submits an
application for the waiver or modification of an
applicable regulation subject to an alternative
compliance strategy.
(2) Appropriate financial regulatory agency.--The term
``appropriate financial regulatory agency'' means--
(A) the appropriate Federal banking agency, as
defined in section 3 of the Federal Deposit Insurance
Act (12 U.S.C. 1813), with respect to an institution
described in subsection (q) of that section;
(B) the Securities and Exchange Commission, with
respect to an institution not described in subparagraph
(A) that is--
(i) any broker or dealer that is registered
with the Commission under the Securities
Exchange Act of 1934 (15 U.S.C. 78a et seq.);
(ii) any investment company that is
registered with the Commission under the
Investment Company Act of 1940 (15 U.S.C. 80a-1
et seq.);
(iii) any investment adviser that is
registered with the Commission under the
Investment Advisers Act of 1940 (15 U.S.C. 80b-
1 et seq.);
(iv) any clearing agency registered with the
Commission under the Securities Exchange Act of
1934 (15 U.S.C. 78a et seq.);
(v) any nationally recognized statistical
rating organization registered with the
Commission under the Securities Exchange Act of
1934 (15 U.S.C. 78a et seq.);
(vi) any transfer agent registered with the
Commission under the Securities Exchange Act of
1934 (15 U.S.C. 78a et seq.);
(vii) any exchange registered as a national
securities exchange with the Commission under
the Securities Exchange Act of 1934 (15 U.S.C.
78a et seq.);
(viii) any national securities association
registered with the Commission under the
Securities Exchange Act of 1934 (15 U.S.C. 78a
et seq.);
(ix) any securities information processor
registered with the Commission under the
Securities Exchange Act of 1934 (15 U.S.C. 78a
et seq.);
(x) the Municipal Securities Rulemaking Board
established under the Securities Exchange Act
of 1934 (15 U.S.C. 78a et seq.);
(xi) the Public Company Accounting Oversight
Board established under the Sarbanes-Oxley Act
of 2002 (15 U.S.C. 7211 et seq.);
(xii) the Securities Investor Protection
Corporation established under the Securities
Investor Protection Act of 1970 (15 U.S.C.
78aaa et seq.); and
(xiii) any security-based swap execution
facility, security-based swap data repository,
security-based swap dealer, or major security-
based swap participant registered with the
Commission under the Securities Exchange Act of
1934 (15 U.S.C. 78a et seq.);
(C) the Bureau of Consumer Financial Protection, with
respect to a covered person, as defined in section 1002
of the Consumer Financial Protection Act of 2010 (12
U.S.C. 5481), that does not have an appropriate
financial regulatory agency under subparagraph (A),
(B), (D), or (E) of this paragraph;
(D) the National Credit Union Administration, with
respect to an insured credit union, as defined in
section 101 of the Federal Credit Union Act (12 U.S.C.
1752); and
(E) the Federal Housing Finance Agency, with respect
to--
(i) a Federal Home Loan Bank;
(ii) the Federal Home Loan Bank System;
(iii) the Federal National Mortgage
Association; and
(iv) the Federal Home Loan Mortgage
Corporation.
(3) Artificial intelligence; ai.--The terms ``artificial
intelligence'' and ``AI'' have the meaning given the term
``artificial intelligence'' in section 5002 of the National
Artificial Intelligence Initiative Act of 2020 (15 U.S.C.
9401).
(4) Commission.--The term ``Commission'' means the Securities
and Exchange Commission.
(5) Federal securities laws.--The term ``Federal securities
laws'' means--
(A) the Securities Act of 1933 (15 U.S.C. 77a et
seq.);
(B) the Securities Exchange Act of 1934 (15 U.S.C.
78a et seq.);
(C) the Sarbanes-Oxley Act of 2002 (15 U.S.C. 7201 et
seq.);
(D) the Trust Indenture Act of 1939 (15 U.S.C. 77aaa
et seq.);
(E) the Investment Company Act of 1940 (15 U.S.C.
80a-1 et seq.);
(F) the Investment Advisers Act of 1940 (15 U.S.C.
80b-1 et seq.);
(G) the Jumpstart Our Business Startup Act (Public
Law 112-106; 126 Stat. 306); and
(H) the Dodd-Frank Wall Street Reform and Consumer
Protection Act (Public Law 111-203; 124 Stat. 1376).
(6) Financial product, service, or activity.--The term
``financial product, service, or activity''--
(A) includes--
(i) a financial product or service, as
defined in section 1002 of the Consumer
Financial Protection Act of 2010 (12 U.S.C.
5481);
(ii) activities that are financial in nature,
as defined in section 4(k)(4) of the Bank
Holding Company Act of 1956 (12 U.S.C.
1843(k)(4));
(iii) any financial product or service
provided by a person regulated by the
Commission, as defined in section 1002 of the
Consumer Financial Protection Act of 2010 (12
U.S.C. 5481); and
(iv) the offer or sale of any security
subject to the Federal securities laws;
(B) includes any internal activity of a regulated
entity that is carried out in furtherance of the
offering or sale of a product, service, or activity
described in subparagraph (A); and
(C) does not include the business of insurance, as
defined in section 1002 of the Consumer Financial
Protection Act of 2010 (12 U.S.C. 5481).
(7) Financial regulatory agency.--The term ``financial
regulatory agency'' means--
(A) the Board of Governors of the Federal Reserve
System;
(B) the Federal Deposit Insurance Corporation;
(C) the Office of the Comptroller of the Currency;
(D) the Securities and Exchange Commission;
(E) the Bureau of Consumer Financial Protection;
(F) the National Credit Union Administration; and
(G) the Federal Housing Finance Agency.
(8) Regulated entity.--The term ``regulated entity'' means an
entity regulated, or insured in the case of a credit union, by
any financial regulatory agency.
SEC. 3. AI INNOVATION LABS.
Each financial regulatory agency shall establish, or identify an
office, division, or department of the agency that shall serve as, an
``AI Innovation Lab'' to enable regulated entities to experiment with
AI test projects without unnecessary or unduly burdensome regulation or
expectation of enforcement actions, including enforcement actions that
disincentivize the substantial use of artificial intelligence to
enhance regulatory compliance practices, pursuant to the approval of an
application under section 4.
SEC. 4. USE OF ARTIFICIAL INTELLIGENCE BY REGULATED FINANCIAL ENTITIES.
(a) Application Submission.--
(1) In general.--Beginning one year after the date of
enactment of this Act, a regulated entity may submit to the
appropriate financial regulatory agency an application, on a
form determined by the appropriate financial regulatory agency,
to engage in an AI test project through an AI Innovation Lab.
(2) Contents.--An application submitted under paragraph (1)
shall include--
(A) a description of the AI test project proposed to
be carried out by the regulated entity, including--
(i) the type and capabilities of the relevant
AI technologies, systems, and applications; and
(ii) the purposes and goals of the AI test
project;
(B) an alternative compliance strategy that--
(i) identifies a regulation issued by a
financial regulatory agency that the regulated
entity requests to be waived or modified;
(ii) proposes an alternative method for the
regulated entity to comply with the Federal
statute under which the regulation identified
in clause (i) is promulgated and the purpose of
such regulation, including an explanation as to
why the alternative method is essential to the
operation of the regulated entity and how the
regulated entity would effectively manage
risks, including cybersecurity risks,
associated with the AI test project;
(iii) describes applicable plans and policies
related to risk assessments, testing, controls,
documentation, and monitoring and their role in
managing risks described under clause (ii); and
(iv) contains policies for the regulated
entity to maintain, consistent with applicable
law, cybersecurity standards, and contractual
limitations and obligations, records of the
activities of the AI test project, so long as
such policies do not require the regulated
entity to maintain any records that the
regulated entity is not permitted to maintain
pursuant to another law or any contractual
limitation or obligation;
(C) an explanation of how under the strategy
described in subparagraph (B), the AI test project--
(i) would serve the public interest, improve
consumer or investor access to a financial
product, service, or activity, or promote
consumer or investor protection;
(ii) would enhance efficiency or operations,
foster innovation or competitiveness, improve
risk management and security, including
cybersecurity, or enhance regulatory
compliance;
(iii) would not present--
(I) a significant risk of loss to the
Deposit Insurance Fund or the National
Credit Union Share Insurance Fund where
the regulated entity is an insured
depository institution or insured
credit union; or
(II) a systemic risk to the financial
system of the United States;
(iv) is consistent with the purposes of the
anti-money laundering and countering the
financing of terrorism obligations under
subchapter II of chapter 53 of title 31, United
States Code; and
(v) would not present a national security
risk to the United States;
(D) a proposed date on which the AI test project
would terminate and an explanation why such termination
date would be appropriate;
(E) proposed limitations on the size, scope, and
growth of the AI test project;
(F) a detailed business plan; and
(G) an estimate of the economic, market, operational,
and public interest impact of the AI test project if
approved.
(3) Joint applications.--Two or more regulated entities may
submit a joint application to the same financial regulatory
agency under paragraph (2).
(4) Regulations of other agencies.--
(A) In general.--Notwithstanding paragraph (2), a
regulated entity may submit an application under this
subsection for an alternative compliance strategy for a
regulation issued or enforced by a financial regulatory
agency that is not the appropriate financial regulatory
agency for the regulated entity.
(B) Requirements.--An application described in
subparagraph (A) shall be subject to the same
requirements as an application described in paragraph
(2), except that--
(i) the regulated entity shall submit the
application to--
(I) the appropriate financial
regulatory agency; and
(II) any other financial regulatory
agency that--
(aa) issued the regulation
that is the subject of the
alternative compliance strategy
and has jurisdiction over the
regulated entity with respect
to such regulation; or
(bb) enforces the regulation
that is the subject of the
alternative compliance strategy
with respect to the regulated
entity; and
(ii) the AI test project may not take effect
unless the application is jointly approved by--
(I) the appropriate financial
regulatory agency; and
(II) each other financial regulatory
agency described in clause (i)(II).
(5) Notice.--A regulated entity that is regulated by, or
subject to regulations that are enforced by, more than 1
financial regulatory agency shall provide notice of any
application submitted to the appropriate financial regulatory
agency under this section to each such financial regulatory
agency not later than 5 business days after the regulated
entity submits the application to the appropriate financial
regulatory agency.
(6) Notice of intervening agency.--
(A) Notice of intervention.--A financial regulatory
agency that receives a notice under paragraph (5) with
respect to an application regarding a regulation th
Purpose and Summary
H.R. 4801, the Unleashing AI Innovation in Financial
Services Act, was introduced on July 29, 2025, by Chairman
French Hill (AR-02).
H.R. 4801 requires financial regulatory agencies to
establish or identify departments to serve as AI Innovation
Labs, enabling regulated entities to test artificial
intelligence (AI) within regulatory sandboxes. The sandboxes
would allow regulated entities to experiment with financial
products, services, and activities under waived or modified
regulations, subject to regulatory approval, terms, conditions,
and limitations, including that the entity implement strategies
for complying with federal law.
Background and Need for Legislation
A sandbox is a regulatory tool that allows regulated
entities to request permission to test a new technology,
product, or service under targeted regulatory relief or
alternative compliance methods. Sandboxes may be subject to
limitations, such as a defined period for the project to run on
a supervised basis, size, scope, and participation caps, and a
potential off-ramp for the project's broader deployment.
Regulatory sandboxes have been an increasingly popular tool
across regulated industries globally, but particularly in
financial services.\1\ Fourteen states have implemented
regulatory sandboxes, many related to financial services.\2\
Sandboxes have been particularly successful in enabling
innovation and providing opportunities for both private sector
innovators and regulators to learn about the development and
application of new technology. The United Kingdom has been
operating a financial regulatory sandbox option since 2014 with
over 800 firms participating and over 160 products accepted for
testing, many in the blockchain and digital asset
industries.\3\
---------------------------------------------------------------------------
\1\Karl E. Schneider, CRS, IF12875, Regulatory Sandboxes at the
Consumer Financial Protection Bureau (2025).
\2\Inst. for Reforming Gov't, Removing Barriers to Innovation: How
Regulatory Sandboxes Can Solve Kitchen Table Issues (2024), https://
reforminggovernment.org/wp-content/uploads/2024/05/
IRG_SandboxReport.pdf.
\3\Innovation Hub: Market Insights, Fin. Conduct Auth. (Mar. 12,
2025), https://www.fca.org.uk/data/innovation-market-insights.
---------------------------------------------------------------------------
As AI continues to reshape industries, including financial
services and housing, it is critical to recognize both the
benefits and risks of this technology. Financial services firms
have long been at the forefront of developing and deploying
machine learning and AI technology. The latest generation of AI
technology, including agentic and generative AI, may pose
questions for existing regulatory frameworks, requiring
adaptation to ensure rules are appropriately tailored. In
addition, these new capabilities may pose novel risks that
financial regulators and industry participants need to address.
A regulatory sandbox for projects making substantial use of AI
at the federal financial regulatory agencies will allow for
productive experimentation under fit-for-purpose supervision,
building both the regulators' and regulated entities'
understanding of the technology and effective governance.
Importantly, in March 2026, the Trump Administration
released its national legislative framework for AI.\4\ The
framework calls for Congress to establish regulatory sandboxes
for AI applications to help unleash American ingenuity and
strengthen U.S. leadership in AI development and deployment.\5\
---------------------------------------------------------------------------
\4\Press Release, The White House, President Donald J. Trump
Unveils National AI Legislative Framework (Mar. 20, 2026), https://
www.whitehouse.gov/releases/2026/03/president-donald-j-trump-unveils-
national-ai-legislative-framework.
\5\Id.
---------------------------------------------------------------------------
Committee Consideration
119TH CONGRESS
On July 29, 2025, Chairman Hill introduced H.R. 4801, the
Unleashing AI Innovation in Financial Services Act, with
Representatives Ritchie Torres (D-NY), Bryan Steil (R-WI), and
Josh Gottheimer (D-NJ) as original cosponsors. The bill was
referred solely to the Committee on Financial Services.
The bill was attached to the September 18, 2025, Digital
Assets, Financial Technology, and Artificial Intelligence
Subcommittee hearing titled, ``Unlocking the Next Generation of
AI in the U.S. Financial System for Consumers, Businesses, and
Competitiveness'' and to the December 10, 2025, Committee
hearing titled, ``From Principles to Policy: Enabling 21st
Century AI Innovation in Financial Services.''
On May 13, 2026, the Committee on Financial Services met in
open session to consider, among others, H.R. 4801. The
Committee ordered H.R. 4801, as amended, to be reported with a
favorable recommendation to the House of Representatives.
On July 29, 2025, Senator Mike Rounds (R-SD) introduced S.
2528, the Unleashing AI Innovation in Financial Services Act, a
Senate companion bill. The bill was referred to the Senate
Committee on Banking, Housing, and Urban Affairs.
118TH CONGRESS
On August 6, 2024, Representative Hill introduced H.R.
9309, the Unleashing AI Innovation in Financial Services Act,
an earlier iteration of H.R. 4801, with Representative Torres
as original cosponsor. The bill was referred solely to the
Committee on Financial Services. There was no further
legislative action for H.R. 9309 in the 118th Congress.
RELATED HEARINGS
Pursuant to clause 3(c)(6) of rule XIII of the Rules of the
House of Representatives, the following hearings were used to
develop H.R. 4801:
On September 18, 2025, the Subcommittee on Digital Assets,
Financial Technology, and Artificial Intelligence held a
hearing titled, ``Unlocking the Next Generation of AI in the
U.S. Financial System for Consumers, Businesses, and
Competitiveness.'' H.R. 4801 was noticed for legislative
consideration at this hearing.
This hearing examined how financial services regulators and
firms are using artificial intelligence in areas such as
lending, fraud detection, and compliance, among others.
Committee Members analyzed both the benefits and potential
risks associated with AI adoption, while assessing the role of
regulators in overseeing its deployment in ways that support
innovation. The Subcommittee heard testimony from: Dr. David
Cox, Vice President, AI Models; IBM Director, MIT IBM Watson AI
Lab; Mr. Christian Lau, Co-Founder and Chief Product Officer,
Dynamo AI; Mr. Matthew Reisman, Director, Privacy and Data
Policy, Center for Information Policy Leadership; Mr. Daniel
Gorfine, Founder & CEO, Gattaca Horizons; Former Chief
Innovation Officer & Director of LabCFTC; and Dr. Nicol Turner
Lee, Senior Fellow and Director, Center for Technology
Innovation, Brookings Institution.
On December 10, 2025, the Committee on Financial Services
held a hearing titled, ``From Principles to Policy: Enabling
21st Century AI Innovation in Financial Services.'' H.R. 4801
was noticed for legislative consideration at this hearing.
This hearing examined artificial intelligence use cases in
the financial services and housing sectors and assessed how
existing laws and regulations apply, identifying where current
frameworks are effective and where gaps may create uncertainty
or hinder innovation. The hearing also explored interagency
coordination to ensure AI oversight is consistent, clear, and
supportive of innovation, competition, and consumer protection.
The Committee heard testimony from: Ms. Jeanette Manfra, Vice
President and Global Head of Risk & Compliance, Google Cloud;
Mr. Tal Cohen, President, Nasdaq; Mr. Nicholas Stevens, Vice
President of Product, Artificial Intelligence, Zillow; Ms.
Wendi Whitmore, Chief Security Intelligence Officer, Palo Alto
Networks; and Mr. Joshua Branch, Big Tech Accountability
Advocate, Public Citizen.
Committee Votes
Clause 3(b) of rule XIII of the Rules of the House of
Representatives requires the Committee Report to include record
votes on the motion to report legislation and amendments
thereto.
On May 13, 2026, the Committee ordered H.R. 4801, as
amended, to be reported favorably to the House by a recorded
vote of 33 yeas and 19 nays, a quorum being present. (Record
Vote No. FC-284).
The Committee considered the following amendments to H.R.
4801:
Representative French Hill (R-AR) offered an
amendment in the nature of a substitute, which requires
entities participating in the sandbox to notify all
applicable federal regulators, and would allow federal
regulators to intervene in sandbox applications
submitted to other agencies when they also supervise
the applicant. The amendment also makes several
technical changes to the application and rulemaking
requirements. This amendment was adopted by a voice
vote.
Representative Maxine Waters (D-CA) offered
an amendment (No. 9), designated HR4801_01. This
amendment delays the effective date by requiring a
study by the Secretary of the Treasury on the effects
of the sandbox framework before the framework is
enacted, as well as on large language models. The
amendment requires that the legislation not take effect
until the Secretary of the Treasury makes a
certification to Congress and the public that the
sandbox framework would not increase risks. This
amendment failed by a recorded vote of 23 yeas and 30
nays, a quorum being present. (Record Vote No. FC-280).
Representative Stephen Lynch (D-MA) offered
an amendment (No. 10), designated LYNCH_058. This
amendment requires applicants for the sandbox to
provide explanations on how the AI test project is
readily explainable to human end users and consumers
and not detrimental to the readability and fairness of
data, decision making, or recommendations. This
amendment failed by a recorded vote of 21 yeas and 32
nays, a quorum being present. (Record Vote No. FC-281).
Representative Lynch offered an amendment
(No. 11), designated LYNCH_060. This amendment requires
applications to the sandbox to include an explanation
of how the AI test project would reduce costs to
consumers. This amendment failed by a recorded vote of
22 yeas and 31 nays, a quorum being present. (Record
Vote No. FC-282).
Representative Lynch offered an amendment
(No. 12), designated LYNCH_061. This amendment adds a
section that states that federal financial regulatory
agencies may utilize all applicable enforcement and
penalty authorities. This amendment failed by a
recorded vote of 23 yeas and 30 nays, a quorum being
present. (Record Vote No. FC-283).
[GRAPHIC(S) NOT AVAILABLE IN TIFF FORMAT]
COMMITTEE OVERSIGHT FINDINGS
Pursuant to clause 3(c) of rule XIII of the Rules of the
House of Representatives, the findings and recommendations of
the Committee, based on oversight activities under clause
2(b)(1) of rule X of the Rules of the House of Representatives
are incorporated in the descriptive portions of this report.
PERFORMANCE GOALS AND OBJECTIVES
Pursuant to clause 3(c)(4) of rule XIII of the Rules of the
House of Representatives, the goal of H.R. 4801 is to enhance
financial innovation by establishing sandbox testing
environments at federal financial regulatory agencies where
regulated entities can develop and test artificial intelligence
projects under tailored regulatory oversight.
COMMITTEE COST ESTIMATE
Clause 3(d)(1) of rule XIII of the Rules of the House of
Representatives requires an estimate and a comparison of the
costs that would be incurred in carrying out H.R. 4801. The
Committee has requested but not received a cost estimate from
the Director of the Congressional Budget Office (CBO). However,
pursuant to clause 3(d)(1) of rule XIII of the Rules of the
House of Representatives, the Committee will adopt as its own
the cost estimate by the Director of the CBO once it has been
prepared.
NEW BUDGET AUTHORITY AND CBO COST ESTIMATE
With respect to the requirements of 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 with respect
to requirements of clause 3(c)(3) of rule XIII of the Rules of
the House of Representatives and section 402 of the
Congressional Budget Act of 1974, the Committee will adopt as
its own the cost estimate for the bill prepared by the Director
of the CBO. However, a cost estimate was not made available to
the Committee in time for the filing of this report. The
Chairman of the Committee shall cause such estimate to be
printed in the Congressional Record upon its receipt by the
Committee.
UNFUNDED MANDATES STATEMENT
The Committee has requested but not received from the
Director of the CBO an estimate of the Federal mandates
pursuant to section 423 of the Unfunded Mandates Reform Act.
The Chairman of the Committee shall cause such estimate to be
printed in the Congressional Record upon its receipt by the
Committee.
EARMARK STATEMENT
In compliance with clause 9 of rule XXI of the Rules of the
House of Representatives, this bill, as reported, contains no
congressional earmarks, limited tax benefits, or limited tariff
benefits as defined in clause 9(e), 9(f), or 9(g) of rule XXI.
FEDERAL ADVISORY COMMITTEE ACT STATEMENT
No advisory committees within the meaning of section 5(b)
of the Federal Advisory Committee Act were created by this
legislation.
APPLICABILITY TO THE LEGISLATIVE BRANCH
The Committee finds that the legislation does not relate to
the terms and conditions of employment or access to public
services or accommodations within the meaning of section
102(b)(3) of the Congressional Accountability Act.
DUPLICATION OF FEDERAL PROGRAMS
Pursuant to clause 3(c)(5) of rule XIII of the Rules of the
House of Representatives, the Committee states that no
provision of the bill establishes or reauthorizes a program of
the Federal Government known to be duplicative of another
Federal program, including any program that was included in a
report to Congress pursuant to section 21 of the Public Law
111-139 or the most recent Catalog of Federal Domestic
Assistance.
SECTION-BY-SECTION ANALYSIS OF THE LEGISLATION
Section 1. Short title
Section 1 provides the short title is the ``Unleashing AI
Innovation in Financial Services Act''.
Section 2. Definitions
Section 2 defines key terms used throughout the bill and
identifies the federal financial regulatory agencies covered by
the Act.
Section 3. AI innovation labs
Section 3 requires the federal financial regulatory
agencies to establish or designate an office, division, or
department to serve as an AI Innovation Lab where regulated
entities may conduct approved AI test projects.
Section 4. Use of artificial intelligence by regulated financial
entities
Section 4 establishes the formal application and approval
process for regulated entities seeking to participate in an AI
Innovation Lab. Applications will become available one year
after enactment and must include information on the AI test
project, the proposed alternative compliance strategy, risk
management and cybersecurity measures, business plans, project
limitations, termination dates, and assessments of the
project's impacts. The section permits joint applicants and
requires entities regulated by multiple federal financial
agencies to notify all applicable regulators, while agencies
with supervisory authority over an applicant to intervene in
the review process. Section 4 requires the agencies to approve
or deny applications within 120 days, subject to a one-time
120-day extension, and directs agencies to issue implementing
regulations within 180 days. The section permits agencies to
seek injunctive relief or issue administrative orders against
projects that violate the terms of their application agreement
or pose risks to consumers or investors, present risks to
financial markets or loss to a Federal deposit or share
insurance fund, or present risks of anti-money laundering
violations and national security risks. The section also
preserves agencies' authority to pursue enforcement actions for
fraud, market manipulation, and unsafe or unsound practices.
Section 5. Annual report
Section 5 requires each financial regulatory agency to
submit annual reports to Congress beginning two years after
enactment and continuing for seven years. Reports must
summarize aggregated findings, trends, and lessons learned from
AI test projects, while excluding the identities of
participating entities and proprietary information. Reports may
include a classified annex.
CHANGES IN EXISTING LAW MADE BY THE BILL, AS REPORTED
H.R. 4801 does not repeal or amend any section of a
statute. Therefore, the Office of Legislative Counsel did not
prepare the report required under clause 3(e) of rule XIII of
the House of Representatives.
[GRAPHIC(S) NOT AVAILABLE IN TIFF FORMAT]
Minority Views
At a time of unprecedented AI risk, H.R. 4801 would give
companies free rein to experiment on consumers and investors on
their own terms. This legislation directs the Fed, FDIC, OCC,
SEC, CFPB, NCUA, and FHFA to each establish an AI Innovation
Lab that would allow companies to experiment with AI projects
without complying with certain regulations or being subject to
enforcement actions. The bill allows entities to propose their
own alternative method to comply with regulations that they
want waived or modified. All the company must do is ``explain''
how their AI project would meet certain characteristics.
However, there is no way for the regulators to look under the
hood and make sure these statements are accurate. Instead,
there is presumed approval by the regulators with no
opportunity for public comment.
Furthermore, this bill would also authorize the private
sector entity, and not the regulator, to set the timeline,
size, scope, and growth of the project. The definition of the
project is broad, allowing these entities to waive or modify
regulations on a wide range of activities, as long as the
project ``makes substantial use of AI.'' Under this bill for
example, companies could waive anti-discrimination laws and
regulations, including the Equal Credit Opportunity Act, the
Home Mortgage Disclosure Act, the Fair Housing Act, and the
Community Reinvestment Act. There also is no restriction on
companies waiving key investor protections, like the fiduciary
duty requirement of the Investment Adviser Act of 1940, which
guards against mismanagement by financial advisors, or market
manipulation laws and regulations. Additionally, there are no
limitations or safeguards in H.R. 4801 that protect consumers,
small businesses, or financial markets from harm.
To make matters worse, this legislation fundamentally
ignores the fears and concerns the American public has in light
of the latest AI developments. In recent years, companies have
integrated Generative AI (Gen AI) and agentic AI (AAI or AI
agents) into various facets of American's lives.\1\ While AI
technology may have some benefits, the American public is
concerned about the scale of its risks. In fact, voters
overwhelmingly support the regulation of AI. Polling shows that
97% of Americans think that, ``AI safety and security should be
subject to rules and regulations.''\2\ Increasing regulation of
AI has bipartisan support--77% of voters, including 75% of
Republicans, say ``government should do more to regulate
AI.''\3\ This bill does the exact opposite. Furthermore, this
legislation does not grapple with AI's current impact on the
workforce and its potential for large-scale disruption. As
companies apply for the regulatory waivers available pursuant
to H.R. 4801 and widely launch their AI projects, these
companies have no requirement test the impact of the project on
the affected workers or on the broader labor market.
Additionally, this bill is silent on the environmental
consequences and harmful health implications that AI data
centers are having on local communities.
---------------------------------------------------------------------------
\1\Gen AI goes beyond analyzing data, and when prompted can create
original content such as text, images, videos, music, and more. See
What is generative AI?, IBM (Apr. 20, 2023). Although Gen AI and AAI
both rely on LLMs as a foundation, Gen AI technology focuses on
creating new content based on learned patterns, while AAI or AI agents
go further by autonomously and proactively executing complex tasks.
See, FinRegLab, The Next Wave Arrives: Agentic AI in Financial Services
(Sept. 2025); and IBM, What is agentic AI? (accessed Sept. 14, 2025).
\2\Years of Polling Show Overwhelming Voter Support for a Crackdown
on AI, Public Citizen (Nov. 21, 2025).
\3\Id.
---------------------------------------------------------------------------
The challenges from Gen AI and AAI are further exacerbated
as AI model capabilities grow exponentially. On April 7,
Anthropic announced information on its Claude Mythos Preview, a
new general purpose language model, which the company describes
as ``extremely autonomous'' and extraordinarily capable at
computer security tasks, software coding, and complex reasoning
objectives.\4\ According to Anthropic, Claude Mythos Preview
``has already found thousands of high-severity vulnerabilities,
including some in every major operating system and web browser.
. . . The fallout-for economies, public safety, and national
security--could be severe.''\5\ These warnings were followed by
urgent and unprecedented meetings between Treasury Secretary
Scott Bessent, Fed Chair Jerome Powell and the CEOs of nation's
largest banks\6\ and meetings between Anthropic and the White
House to discuss AI safety and risks.\7\ World leaders and
regulators continue to express concerns regarding the scale and
speed of Mythos' and other AI models' capabilities.\8\ Yet
surprisingly, this bill contains no reference to these
developments and instead could threaten our financial system's
safety and soundness at a time when Congress' focus should be
on security.
---------------------------------------------------------------------------
\4\Anthropic holds Mythos model due to hacking AXIOS (Apr. 7,
2026); Anthropic, Project Glasswing (accessed Apr. 24, 2026).
\5\Anthropic, Project Glasswing (accessed Apr. 24, 2026).
\6\Bessent, Powell warned bank CEOs about Anthropic model risks,
sources say, Reuters (Apr. 9, 2026); Bessent and Powell's A.I. Anxiety,
New York Times (Apr. 10, 2026).
\7\Anthropic's New A.I. Model Sets Off Global Alarms, New York
Times (Apr. 22, 2026).
\8\Global regulators trail banks in AI as Mythos raises oversight
concerns, report finds, Reuters (Apr. 28, 2026).
---------------------------------------------------------------------------
During the markup, Democrats offered four amendments aimed
at alleviating the deficiencies within the legislation, all of
which were rejected by Republicans:\9\
---------------------------------------------------------------------------
\9\House Committee on Financial Services, Committee Markup, 19th
Cong. (May 13, 2026).
---------------------------------------------------------------------------
Ranking Member Waters' amendment would
prevent the bill from taking effect until there is a
study by Treasury, in consultation with the other
regulators in the bill, that confirms that the bill
would not increase risks to consumers from models like
Mythos;
Rep. Lynch's first amendment adds two new
criteria that an A.I. system must meet to qualify for a
test project: (1) it must be explainable to users and
consumers; and (2) must not have unreliable or unfair
outputs;
Rep. Lynch's second amendment requires
applicants to explain how their A.I. test project will
reduce consumer costs as a condition of approval, and
report this information to the appropriate agency; and,
Rep. Lynch's third amendment clarifies that
financial regulatory agencies retain full enforcement
and penalty authority, if a person is harmed as a
result of an AI test project.
H.R. 4801 is opposed by 90 civil rights, consumer, labor,
technology, and other advocacy groups.\10\
---------------------------------------------------------------------------
\10\Americans for Financial Reform, Letters to Congress: AFR Joined
by 89 Groups to Oppose Financial AI Sandbox Giveaway (May 12, 2026).
---------------------------------------------------------------------------
For all of these reasons, we oppose H.R. 4801.
Maxine Waters,
Ranking Member.
Nydia M. Velazquez,
Bill Foster,
Al Green,
Emanuel Cleaver, II,
Joyce Beatty,
Ayanna Pressley,
Rashida Tlaib,
Members of Congress.
[all]