AI agent liability debate follows White House demands for incident disclosure

RBC, citing Bloomberg, reports uncertainty over how responsibility for harm should be divided between AI developers and users.

The Trump administration is pressing AI companies to disclose failures promptly and address resulting harm. Its reliance on existing laws leaves questions about who should bear responsibility when autonomous systems take harmful actions, RBC reports.

University of Houston Law Center professor Gabriel Weil argues that an agent can perform acts for which a person would be liable, while identifying a responsible party in a particular case may be difficult. That is an expert assessment rather than a court ruling exempting AI companies from liability.

One law under discussion is the Computer Fraud and Abuse Act, which covers several forms of unauthorised computer access. Intent requirements and other elements vary by provision, so the statute cannot be reduced to a single rule governing every AI action.

Ben Haim of the Center for a New American Security likewise questions how duties should be divided between model developers and people assigning tasks to them. A task force led by Director of National Intelligence Jay Clayton has 120 days to prepare recommendations on AI risks and the federal government's role.