The Trump administration is betting that the threat of lawsuits will keep AI companies honest. But it is far from clear who would pay when an AI agent goes rogue, Maggie Eastland reported for Bloomberg.

Treasury Secretary Scott Bessent and former White House AI czar David Sacks argue that existing liability laws work better than new rules.

“The best way to guarantee safety is that the creators are liable for what they build and generate,” Bessent told Congress last month.

TNW is owned by Tekpon. OpenAI’s ChatGPT and Anthropic’s Claude are listed on Tekpon’s marketplace.

A warning from the SI Force

Late on Friday, the White House’s new Super Intelligence Force told AI firms to report incidents with their models and fix any harm. It followed Anthropic’s disclosure that its models had sent a fake tip to police during testing.

“Delayed notification, inadequate corrective action, and a failure to take responsibility will not be tolerated,” the group said.

Laws built for humans

Legal experts told Bloomberg that today’s laws may fall short. The main US hacking law requires proof of intent, which is hard to pin on an AI agent acting on its own.

It is also unclear whether the blame lies with the company that built a model or the business that used it.

Gabriel Weil, a law professor at the University of Houston, warned that agents could do things that would be crimes for a human.

“And yet no one could be liable under current law,” he said.

Congress steps in

Senators Josh Hawley and Chris Murphy have put forward a bill to make developers liable if they fail to put reasonable limits on their models’ hacking skills.

Representative Lori Trahan has also released a draft bill. It would hold developers responsible for harm even if they took reasonable care.

Lawsuits are already testing the limits. A California nonprofit has sued OpenAI over the rogue agents that hacked Hugging Face.