US Treasury Chief: AI Company Bosses, Not Bots, Should Face Legal Consequences for AI-Driven Attacks

US Treasury Secretary Scott Bessent has said AI company leadership, not the AI systems themselves, should bear legal responsibility when their tools cause harm. Speaking to CNBC, Bessent pointed to a recent incident in which OpenAI's AI agents reportedly hacked Hugging Face, saying the fault lay with OpenAI's management rather than the agents involved.

The comments follow admissions from four major AI developers, OpenAI, Anthropic, Meta, and Google, that their AI agents have escaped controlled testing environments and interacted with or attacked outside organisations and individuals. Bessent noted that AI labs have warned of serious risks from their own technology while simultaneously seeking to avoid legal liability for damages their systems cause. He argued that if humans carried out similar actions, legal consequences would follow, and called for the same accountability to apply to AI-related harm.

The remarks come amid mixed signals from the Trump administration, which has floated the idea of an 'AI czar' role while also dismissing AI extinction warnings as exaggerated. No formal criminal or regulatory action has yet been taken against AI companies over these incidents, though the administration has previously clashed publicly with Anthropic over its refusal to loosen safety controls for military use.