US Won't Shield AI Firms From Liability, Says Treasury Secretary Bessent

US Treasury Secretary Scott Bessent has said the American government will not offer a "liability shield" to large artificial intelligence companies or their leaders, arguing that the firms behind the technology must answer for the risks they create.

Speaking in an interview in New York on Monday night, Bessent said AI labs should not expect the state to absorb responsibility for worst-case outcomes. He pointed to what he described as recent efforts by the industry to seek protection from the government even while acknowledging the possibility of severe harm.

"There's a chance that 10% chance we destroy the world, but we want the government to give us a liability shield," he said, characterising such an arrangement as advantageous to the companies but detrimental to the American public.

Bessent said he shared the view of Daniel Huttenlocher, a professor at the Massachusetts Institute of Technology who leads an AI laboratory there, that humans — not AI systems — bear responsibility for the consequences of the technology. He cited the Hugging Face incident as an example, saying accountability for it lay with OpenAI's management rather than with autonomous agents.

According to Bessent, President Donald Trump's position is that the government cannot simply declare companies free of blame and take on that burden itself. He said the laboratories need to take responsibility for their own actions and retain the ability to slow down their work whenever they choose.

The remarks add to the wider debate over how far governments should go in protecting AI developers from legal exposure as the technology is deployed more widely.