United States of America: Senator released draft of TRUMP AMERICA AI Act including testing requirement

On 18 March 2026, a United States Senator released a discussion draft of The Republic Unifying Meritocratic Performance Advancing Machine intelligence by Eliminating Regulatory Interstate Chaos Across American Industry Act. The Act would require each developer of an advanced AI system, defined as a system trained using more than 10²⁶ computing operations, to participate in the Advanced Artificial Intelligence Evaluation Programme and to provide the Secretary of Energy, on request, with materials including the underlying code, training data, model weights, the interface engine, and detailed information regarding training and model architecture. No person would be permitted to deploy an advanced AI system for use in interstate or foreign commerce unless in compliance with these obligations, with a penalty of not less than USD 1'000'000 per day of violation. The Secretary of Energy would be required to establish the Programme within 90 days of enactment. The Programme would conduct standardised and classified testing and adversarial red-team testing at a level matching sophisticated malicious actors, facilitate independent third-party blind model evaluations, and provide participating entities with formal reports identifying evaluated risks. The Programme would terminate 7 years after enactment unless renewed by Congress. The Act would further require each provider of a high-risk AI system to subject that system to an annual independent third-party audit to detect viewpoint discrimination or discrimination based on political affiliation, with a report submitted to the FTC not later than 180 days after completion of each audit. Each covered entity would also be required to provide all personnel with annual AI ethics training using a curriculum established by the FTC. The Act would additionally require NIST to support the development of voluntary, consensus-based testing standards for AI system components and to establish AI blue-teaming capabilities, and would require the Under Secretary of Commerce and Secretary of Energy to jointly establish a testbed programme within 1 year of enactment for voluntary testing, evaluation, and security risk assessment of AI systems, including a voluntary foundation model test programme. The Under Secretary of Commerce would further be required to establish a public-private partnership to develop guidelines, metrics, and practices for evaluating synthetic content detection tools, including through AI red-teaming and blue-teaming.