S. 1633 (119th)Bill Overview

TEST AI Act of 2025

Science, Technology, Communications|Science, Technology, Communications
Cosponsors
Support
Bipartisan
Introduced
May 7, 2025
Discussions
Bill Text
Current stageCommittee

Read twice and referred to the Committee on Commerce, Science, and Transportation.

Introduced
Committee
Floor
President
Law
Congressional Activities
01 · The brief
Plain-English summaryWhat this bill actually does

This bill requires the Director of NIST to run a pilot program that builds and demonstrates testbeds to develop measurement standards for evaluating artificial intelligence systems used by federal agencies. It mandates a memorandum of understanding with the Department of Energy to access DOE resources, creates an Artificial Intelligence Testing Working Group to draft a strategy, sets timelines for publication and testbed development, allows hiring experts, and requires a post-demonstration report to Congress with recommendations.

Why people may split

Tradeoff: stronger federal standards versus risk of stifling private innovation

Watch point

Relative to its intended legislative type, this bill is a reasonably well-structured administrative/operational measure that establishes authority, sequencing, and interagency coordination to run a pilot program and produce an initial set of testbeds and measurement-standard strategies, but it omits key operational and resourcing details.

This bill requires the Director of NIST to run a pilot program that builds and demonstrates testbeds to develop measurement standards for evaluating artificial intelligence systems used by federal agencies.

It mandates a memorandum of understanding with the Department of Energy to access DOE resources, creates an Artificial Intelligence Testing Working Group to draft a strategy, sets timelines for publication and testbed development, allows hiring experts, and requires a post-demonstration report to Congress with recommendations.

Passage50/100

A narrow, technical bill appealing to agency, industry, and academic stakeholders with limited fiscal demands; success hinges on appropriations and bipartisan agreement.

CredibilityPartially aligned

Relative to its intended legislative type, this bill is a reasonably well-structured administrative/operational measure that establishes authority, sequencing, and interagency coordination to run a pilot program and produce an initial set of testbeds and measurement-standard strategies, but it omits key operational and resourcing details.

Contention56/100

Tradeoff: stronger federal standards versus risk of stifling private innovation

02 · What it does

Who stands to gain, and who may push back.

Likely benefits vs burdens50% / 50%
Federal agenciesFederal agencies · Workers

These are examples from the analysis, not a ranked list of the most-affected groups.

Likely helped
  • Federal agenciesCreates objective measurement standards to improve federal AI procurement, evaluation, and comparability across agencie…
  • Potential benefitStrengthens security and privacy testing to better detect data leakage and vulnerabilities.
  • Federal agenciesPromotes interoperability and comparability of AI system performance across federal programs.
Likely burdened
  • Federal agenciesImposes additional compliance costs on federal agencies and AI vendors to meet new measurement standards.
  • Federal agenciesMay centralize federal standards, potentially displacing state or private standard-setting efforts.
  • WorkersProhibiting citizens of listed countries could limit international collaboration and access to global talent.
03 · Why people split

Why the argument around this bill splits.

Tradeoff: stronger federal standards versus risk of stifling private innovation
Progressive80%

Likely broadly supportive because the bill advances standards for AI reliability, bias, privacy, and transparency, and directs federal coordination.

May seek stronger explicit civil-rights and public-participation safeguards and clearer funding and enforcement mechanisms.

Leans supportive
Centrist75%

Generally favorable as a pragmatic, targeted federal effort to create measurable evaluation tools for AI used by agencies.

Would seek clarity on cost, interagency roles, intellectual property handling, and safeguards against duplication with existing efforts.

Leans supportive
Conservative35%

Cautious to opposed because the bill expands federal involvement in AI testing and standards, potentially increasing regulation and bureaucracy.

May accept narrow, security-focused aspects but worry about innovation impacts and costs.

Likely resistant
04 · Can it pass?

The path through Congress.

Introduced

Reached or meaningfully advanced

Committee

Reached or meaningfully advanced

Floor

Still ahead

President

Still ahead

Law

Still ahead

Passage likelihood50/100

A narrow, technical bill appealing to agency, industry, and academic stakeholders with limited fiscal demands; success hinges on appropriations and bipartisan agreement.

Scope and complexity
24%
Scopenarrow
52%
Complexitymedium
Why this could stall
  • No explicit authorization of appropriations included
  • Overlap with existing NIST or DOE programs and authorities
05 · Recent votes

Recent votes on the bill.

No vote history yet

The bill has not accumulated any surfaced votes yet.

06 · Go deeper

Go deeper than the headline read.

Included on this page

Tradeoff: stronger federal standards versus risk of stifling private innovation

A narrow, technical bill appealing to agency, industry, and academic stakeholders with limited fiscal demands; success hinges on appropriat…

Unlocked analysis

Relative to its intended legislative type, this bill is a reasonably well-structured administrative/operational measure that establishes authority, sequencing, and interagency coordination to run a pilot program and pro…

Go beyond the headline summary with full stakeholder mapping, legislative design analysis, passage barriers, and lens-by-lens tradeoff breakdowns.

Perspective breakdownsPassage barriersLegislative design reviewStakeholder impact map
Open full analysis