TEST AI Act of 2025
- Bill Number
- S. 1633
- Origin Chamber
- Senate
- Congress
- 119th Congress, Session 1
- Policy Area
- Science, Technology, Communications
- Status
- Introduced
- Latest Action
- 2025-05-07: Read twice and referred to the Committee on Commerce, Science, and Transportation.
- Last Updated
- 2025-06-10T14:38:07Z
AI-Generated Summary
Purpose
The legislation, titled the "Testing and Evaluation Systems for Trusted Artificial Intelligence Act of 2025" (or "TEST AI Act of 2025"), aims to promote trustworthy artificial intelligence (AI) by directing the National Institute of Standards and Technology (NIST) to create a pilot program. This program uses specialized testing facilities, called testbeds, to develop reliable measurement standards for evaluating AI systems, particularly those used by federal government agencies. The goal is to ensure AI is assessed for qualities like reliability, security, and fairness through transparent and repeatable methods.
Key Provisions
- Definitions:
- "Artificial intelligence system" refers to AI as defined in the 2020 National Artificial Intelligence Act (essentially, technology that simulates human intelligence for tasks like decision-making).
- "Covered foreign country" includes China, Russia, North Korea, and Iran (relevant for restricting participation in advisory groups).
- "Testbed" is a facility or setup for rigorous, transparent testing of technologies to check their function, performance, and security.
- Other terms define NIST's Director and a new advisory "Working Group."
- Pilot Program Establishment:
- NIST's Director must launch the program in coordination with the Department of Energy (DOE) and after consulting the Working Group.
- Focus: Test the feasibility of creating measurement standards for AI evaluation in federal use, using an iterative process where testing advances science, develops standards, and involves reviews by government, private sector, and academic stakeholders.
- Memorandum of Understanding (MOU):
- Within 180 days of enactment, the Secretaries of Commerce (overseeing NIST) and Energy must sign an MOU to enable coordination, including NIST's access to DOE resources like personnel, facilities, and research programs.
- The MOU must be reassessed every 2 years starting 2 years after the pilot begins, to ensure it supports ongoing needs.
- Artificial Intelligence Testing Working Group:
- Established within 90 days of the MOU by the Secretary of Commerce; limited to 10 members, including representatives from Commerce, Energy, NIST, private sector, and academia.
- Prohibition: No members can be citizens of covered foreign countries.
- Duties: Within 1 year, develop and publish (on a public Commerce website) a strategy for AI measurement standards. The strategy must cover:
- Needed standards for AI reliability, performance, capability, interpretability (how decisions are explained), security (e.g., data leak risks), privacy, data bias (unfair treatment based on flawed inputs), and use case limits.
- A blueprint for developing these standards.
- Initial target AI applications.
- Metrics to evaluate the pilot program itself.
- Purpose: Advise Commerce and Energy on standards development.
- Testbed Development:
- Within 2 years of enactment, NIST's Director, with DOE and Working Group input, must build testbeds based on the strategy to demonstrate AI measurement standards for federal agencies.
- NIST can hire experts from academia, industry, or standards organizations as needed.
- Reporting Requirement:
- Within 180 days after the first testbed demonstration, NIST must report to Congress on findings, recommended pilot revisions (including resource needs), and suggestions for laws or administrative actions to further AI standards.
Significant Changes to Existing Law
This bill introduces a new pilot program and advisory structure not previously mandated by law. It builds on the 2020 National Artificial Intelligence Act by adopting its AI definition but adds specific requirements for testbeds and measurement standards. No direct amendments to existing statutes are made; instead, it creates fresh mechanisms for AI evaluation, emphasizing federal coordination and stakeholder involvement.
Potential Impacts
- Government Agencies: Federal entities using AI (e.g., in defense, healthcare, or research) could benefit from standardized evaluation tools, leading to more secure and effective AI deployment. NIST and DOE would gain enhanced collaboration, potentially requiring new funding or personnel.
- Citizens: Improved AI standards may reduce risks like biased decisions or privacy breaches in government AI applications (e.g., benefits processing or surveillance), fostering public trust in technology.
- International Relations: By barring citizens from certain adversarial nations (China, Russia, North Korea, Iran) from the Working Group, the bill could signal U.S. priorities in AI security amid global competition, possibly straining ties with those countries but strengthening alliances with like-minded partners.
Main Stakeholders Affected
- Government: NIST (lead agency), Department of Commerce, Department of Energy, and other federal agencies relying on AI.
- Private Sector: AI developers, tech companies, and standards organizations involved in testing and standards creation.
- Academia and Research: Universities and experts contributing to the Working Group and testbed development.
- Congress: Receives reports and recommendations, influencing future AI policy.
- Broader AI Community: Developers, researchers, and users who can access the published strategy for guidance.
Notable Legal, Constitutional, or Political Implications
- Legal: Establishes enforceable timelines (e.g., 180 days for MOU, 2 years for testbeds) and a formal Working Group, potentially setting precedents for inter-agency AI collaboration. The foreign citizen prohibition could raise questions about equal participation but aligns with national security laws.
- Constitutional: No direct challenges; supports the government's role in promoting science and technology under the Commerce Clause, without infringing on free speech or due process.
- Political: Bipartisan sponsorship (e.g., Senators Lujan, Blackburn, Durbin) highlights cross-party consensus on AI trustworthiness. It addresses U.S. leadership in AI amid geopolitical tensions, but implementation may depend on funding appropriations, which could become partisan.
This summary was generated by AI and may contain inaccuracies. Refer to the official source document for the authoritative text.
Sponsor
Cosponsors (4)
Sen. Blackburn, Marsha [R-TN], Sen. Durbin, Richard J. [D-IL], Sen. Risch, James E. [R-ID], Sen. Welch, Peter [D-VT]
Recent Actions
- 2025-05-07: Read twice and referred to the Committee on Commerce, Science, and Transportation.
- 2025-05-07: Introduced in Senate
Bill Versions
- Testing and Evaluation Systems for Trusted Artificial Intelligence Act of 2025 — issued 2025-05-07 — PDF (9 pages)