Treasury Secretary Bessent: Congress Should Not Give AI Labs Liability Immunity
The federal liability shield the AI industry has fought for is, according to new reporting, genuinely off the table — not killed by Congress, but by Treasury Secretary Scott Bessent, who believes "the best way to guarantee safety is that the creators are liable for what they build." At the same time, he has compared regulation-friendly AI executives to Hannibal Lecter.
The frontier labs' most ambitious political play this year — a federal liability shield meant to protect the laboratories from lawsuits when their models cause harm — is in practice dead. According to reporting from Implicator.ai, republished by Startup Fortune, it was not Congress that closed the door, but Treasury Secretary Scott Bessent: first on CNBC's "Squawk Box" on September 21, where he said AI developers must stop looking to Washington for protection, and then in his testimony before the House Financial Services Committee, where he argued that "the best way to guarantee safety is that the creators are liable for what they build and generate" — and that Congress should not give AI labs any liability exemption.
The statements have concrete content. According to the Implicator.ai reporting, the shield is "effectively shuttered" — not formally withdrawn, but politically without support from the administration's central finance minister. For an industry that has pushed for legal protection against model-harm lawsuits, that is a real political outcome.
The Administration Is United Against the Shield
Bessent is not alone. According to the same reporting, David Sacks, the White House's AI chief, has called the self-regulation proposals built around the shield a potential case of regulatory capture — the industry capturing regulation for its own protection. FTC Chair Andrew Ferguson said that requests combining new regulation with an antitrust exemption trigger "all of my alarm bells."
The context is a meeting President Donald Trump arranged at the White House with AI executives, which resulted in a voluntary self-policing pledge. The administration's line is thus consistent: no new federal regulation, no liability exemption — voluntary self-policing instead.
"Hannibal Lecter: Stop Me Before I Kill Again"
Bessent has also attacked those in the industry who are calling for regulation. In an Axios interview published Saturday, October 3 — which Bloomberg independently covered — he compared regulation-friendly AI executives to the fictional cannibal:
"I mean, it's kind of like Hannibal Lecter: Stop me before I kill again," Bessent said, according to Yahoo/Business Insider. "This alarmism without solutions by some of the AI community — that's not leadership."
Bloomberg summarized the core message: Bessent criticized prominent AI figures' warnings about existential risks as alarmist and unhelpful, and urged the industry to police itself and deliver solutions.
He also pointed to what he sees as evidence that it works. Bessent cited OpenAI's recent decision to shelve a new model after internal tests raised questions about whether the model would follow users' instructions. "The labs have been very cooperative," he said.
The Backdrop: Amodei's Slowdown Push
The interview came after Anthropic CEO Dario Amodei warned in September that "we must slow down the pace" of AI development to prevent the technology from running out of control and harming people. Elon Musk and OpenAI CEO Sam Altman — two rivals who rarely agree publicly — quickly endorsed the call, according to Mediaite via Yahoo. Trump himself has dismissed doomsday concerns as a hoax and promised no new regulations.
But the self-policing line has a structural weakness: According to the Implicator.ai reporting, a pacing framework of Amodei's type in practice needs an antitrust exemption to function without the participating labs facing collusion charges. Coordinated slowdown between competitors is exactly what US antitrust law is designed to catch — without waivers or legislation, voluntary mutual restraint is legally dangerous terrain.
Where It Goes Next
The liability debate does not disappear with the shield. Senators Josh Hawley and Chris Murphy are preparing AI liability legislation, according to Axios — a sign that skepticism of both self-regulation and liability exemptions has bipartisan footing far beyond the Treasury Department. If such bills become reality, the labs will face precisely the accountability mechanism they tried to avoid via the shield — this time without the administration's protection.
Caveats
Part of the basis rests on secondary sources. The core claim that the shield is shut down comes from Implicator.ai via Startup Fortune, which in turn builds on CNBC. There is no official confirmation of whether the shield has been formally withdrawn, shelved, or merely opposed. Startup Fortune also describes an incident in which OpenAI's AI agents were linked to a security breach at Hugging Face, with Bessent blaming the people who run the technology — but the breach itself is not independently documented in the available source material, and is reported here only as Startup Fortune's rendition of CNBC's transcript. These uncertainties do not change the main point: The administration's position on who pays when AI causes harm is now concrete — and it points at the labs themselves.