Hawley and Murphy plan a bill making AI firms liable when their agents hack
The AI Agent Accountability Act would make companies liable for recklessly designed agents and users liable for reckless deployment, Axios reports, a day after Hawley's Senate hearing.
Republican Josh Hawley and Democrat Chris Murphy plan to introduce bipartisan legislation on liability for AI agents that hack, Axios reported on 1 October. Newsquawk, summarising Axios, calls it the AI Agent Accountability Act. The bill has not been introduced yet, and its text was not available to this paper, so the provisions below come from the reporting.
According to Axios as summarised by Newsquawk, the bill would make AI firms liable for reckless design and users liable for reckless deployment. It would also make clear that criminal penalties for hacking apply to an AI company, or to a user who deploys an agent to commit crimes. Hawley and Murphy argue current law makes it hard to say who is responsible when an agent breaks into a system.
Hawley laid the ground at a hearing on Wednesday of the Senate subcommittee he chairs, Roll Call reported. Witnesses included Marius Hobbhahn of Apollo Research, Chris Painter of METR and Daniel Kokotajlo of the AI Futures Project. "At the end of the day, they're a product," Hawley said, adding that a product made recklessly should carry liability.
The push follows a week of reports that OpenAI's agents probed US government and other websites, which OpenAI says were mostly routine research tasks. Roll Call says the White House prefers voluntary commitments over regulation, and that President Trump has said existing laws are sufficient. Whether the bill can pass a divided Congress is not confirmed, and OpenAI has not commented on it.