U.S. Congressman unveils the outline of the SAFE AI Act, proposing to establish federal safety standards for cutting-edge AI
On November 19, U.S. Representatives Don Beyer (Democrat, Virginia), Sara Jacobs (Democrat, California), and Celeste Maloy (Republican, Utah) released the outline of the "Safe and Accountable Federal Assessment of Artificial Intelligence Act" (SAFE AI Act). This outline corresponds to bipartisan legislation that has not yet been formally submitted, aiming to establish federal safety requirements for frontier artificial intelligence and requiring frontier laboratories to develop safety frameworks, report safety incidents, and enhance transparency, while also providing for government oversight and due process.
The outline proposes that the Department of Commerce's Center for Artificial Intelligence Standards and Innovation (CAISI) will work with the intelligence community, the Department of Energy, and other agencies to establish minimum requirements for safety frameworks. Frontier developers must publish and adhere to enforceable safety frameworks and cannot unilaterally lower standards; pre-deployment assessments will be conducted by government teams and government-certified third parties, and developers cannot pay the assessors directly. The proposal also includes civil penalties, emergency authority to order mitigation measures in the presence of an imminent risk of serious harm, judicial review, whistleblower protection, and aims to clarify developers' liability for reasonably foreseeable harm to third parties. The outline positions federal rules as minimum standards, allowing states to establish stricter requirements; startups and academic institutions are generally not covered.