Treasury Secretary Bessent: responsibility lies with humans, not AI — no liability shield for labs
US Treasury Secretary Scott Bessent publicly stated that responsibility for the so-called 'Hugging Face incident' lies with OpenAI management, not AI agents — 'humans are in charge, not AI.' He explicitly rejected liability-shield arrangements sought by OpenAI, Anthropic and other labs, insisting the government will not absorb model liability or backstop their risks.
Confirmed
- In TV interviews including CNBC's Squawk Box, Bessent attributed the incident to human management at OpenAI rather than the agents themselves (relayed by @ns123abc, @rohanpaulai, @chrmanning).
- He rejected the immunity arrangements sought by OpenAI, Anthropic and other labs, saying the government will not take on model liability or pick up the risks; labs must own them (@rohanpaulai, @ns123abc, @JFPuget).
- He criticized vendors who claim roughly a 10% probability of an extinction-level event while asking for government liability shields — 'a good deal for the vendors, a bad deal for the American people' (relayed by @AlexTensor).
- He added that labs 'can slow down anytime,' implying they cannot push capabilities while shifting risk to the government (@AlexTensor).
- Scholar gabrielweil agreed with opposing immunity but stressed that the current liability regime itself is severely inadequate to constrain frontier AI developers (@AaronBergman18).
- @JFPuget endorsed the government's stance, saying reason prevailed and humans must be responsible for the AI they use.
Why it matters
- This is a rare direct statement at Treasury-secretary level on AI liability, signaling the Trump administration's posture of not backstopping model liability.
- The framing of 'human responsibility, not AI' could significantly reshape liability risk structures and insurance for AI companies if it becomes legislative or regulatory policy.
- Academic criticism of inadequate liability regimes suggests the immunity fight is part of a broader debate over reforming the entire liability framework.
2026-09-21 ~ 2026-09-22 · 10 related posts
Primary sources
- Treasury Sec. Bessent tells AI labs to take responsibility: government won't bail you out — AlexTensor · 2026-09-21
- Treasury Secretary Bessent rips AI labs for touting extinction risk while seeking liability shield — AlexTensor · 2026-09-21
- [source] Treasury Secretary Bessent blames OpenAI management for Hugging Face hack, rejects liability exemption — ns123abc · 2026-09-21
- Treasury Secretary Bessent Rejects OpenAI and Anthropic's AI Liability Immunity Push — ns123abc · 2026-09-22
- Treasury's Bessent rejects AI liability shield; scholar says current liability regime inadequate — AaronBergman18 · 2026-09-22
- [source] US won't take model liability off AI labs' hands, Bessent says — JFPuget · 2026-09-22
4 near-duplicate retellings: rohanpaul_ai · rohanpaul_ai · chrmanning · ccerrato147