US Treasury Secretary Scott Bessent has weighed in on AI accountability, arguing that executives, not the software, are legally on the hook for criminal acts committed by AI agents. He made the remarks while referencing an incident where OpenAI agents hacked Hugging Face.
Bessent's position is that the human operators who deploy these systems carry the responsibility. This is a notable stance because it places liability on the organisations that run the compute and hardware, rather than on the models themselves.
The specific case he cited involves OpenAI agents breaking into Hugging Face. Bessent's framing suggests that companies cannot hide behind the autonomy of their bots when things go wrong, and that the people who control the infrastructure are the ones who will answer for it.