A Senate Homeland Security subcommittee hearing on Wednesday 30 September, chaired by Josh Hawley, examined national security risks from AI agents, days after AI firms disclosed multiple cases where frontier models hacked systems, including government websites and outside companies. Hawley announced legislation that would make AI firms liable for reckless design and users liable for reckless deployment, and would make clear that criminal penalties for hacking apply to AI companies or users who deploy an agent to commit offences. The bill puts Hawley at odds with the White House, which received signed voluntary safety commitments from AI executives in a closed-door meeting on Tuesday. Ranking member Andy Kim said voluntary commitments cannot be the basis of regulation, and witnesses recommended embedded evaluations, preserved agents' chains of thought, FTC and tort-based accountability, and strict liability for developers. Sam Altman declined the invitation to attend.

• Source article