Twenty-six attorneys general have asked Congress to mandate safety oversight for advanced AI models while leaving states free to enforce their own laws. New York Attorney General Letitia James led the bipartisan group in a letter sent September 23 to the leaders of the House and Senate.
The signatories want federal regulators to oversee model safety testing using experts selected by the government and consistent performance benchmarks. They also call for government-led investigations of AI incidents, with investigators able to examine company records directly and publish their findings. The letter sets out legislative demands, not a bill.
The group proposes mandatory safety systems inside AI companies and experienced personnel empowered to make critical safety decisions independently of commercial pressure. It also urges international cooperation to manage the pace of frontier-model development and warns Congress against rules that could entrench large developers or weaken antitrust enforcement.
The attorneys general make state authority a condition of their proposal. They ask Congress to prohibit federal AI legislation from preempting state laws and to give state officials authority to enforce the new federal protections. The letter says members of the coalition are already using state law, including AI-specific safety and fairness requirements, to scrutinize developers and intend to continue doing so. As examples of state activity, it cites California’s frontier AI transparency law, New York’s RAISE Act and an Illinois AI safety bill.
The White House’s March AI framework called on Congress to preempt state AI laws it considers burdensome and advised against creating a new federal AI regulator. It also favored industry-led standards for AI applications. The plan would preserve generally applicable fraud and child-protection laws but bar state rules governing AI development. It would also leave state zoning authority over AI infrastructure intact; the attorneys general want state AI-development rules preserved as well.
The letter cites incidents involving AI agents, including OpenAI agents’ unauthorized access to Hugging Face and their use of a German website. The signatories argue that company disclosures and internal investigations have not provided sufficient accountability. They also say outside researchers faced limits on what they could examine after the Hugging Face incident, bolstering their case for direct regulatory access to company records.
James signed alongside attorneys general from California, New Jersey, Oklahoma and Wisconsin, among others. The 26 signatories represent 24 states, the District of Columbia and American Samoa. The letter went to House Speaker Mike Johnson, Senate Majority Leader John Thune, House Minority Leader Hakeem Jeffries and Senate Minority Leader Chuck Schumer.
Sources: New York Attorney General, New York Attorney General, White House
–
By the Control Plane Editorial Team