Senate proposes AI regulation bill requiring companies to demonstrate risk mitigation measures

The U.S. Senate is considering a new AI regulation bill that would require companies to prove they have implemented reasonable measures to prevent harm from their AI tools. According to a source familiar with the negotiations, the bill would grant the Secretary of Commerce the authority to demand evidence from AI developers that they have fulfilled their duty of care. This includes the power to deploy government auditors to test products directly. The proposal is still in the negotiation phase and faces significant hurdles before becoming law. The debate has intensified following Jacob Cockayne’s resignation from Anthropic, where he cited concerns that AI could potentially end human life by the 2030s. In response, major AI firms have called for a slowdown in development, leading to a decline in global AI stocks on September 14. The bill also includes provisions that would prevent states from enforcing their own AI-related regulations, with further details yet to be finalized.