Bessent told the House not to give AI labs a liability blank check
Treasury Secretary Scott Bessent told the House Financial Services Committee on Tuesday that the government should not give frontier AI labs a liability exemption, Matt Bracken at FedScoop reports. Asked by Rep. Juan Vargas what the administration is doing on AI safety, Bessent said Treasury has been “working on safety nonstop since the release of Mythos.”
Mythos is Anthropic’s model whose cybersecurity risks prompted an April meeting at Treasury with then-Fed Chair Jerome Powell and bank CEOs. A liability exemption is a legal shield so a company can’t be sued, or is hard to sue, when its product causes harm. An open-source model is AI whose weights and code others can inspect and run.
“What [government] shouldn’t do on safety is to give these labs a liability exemption, which is what they are asking for,” Bessent said. “The best way to guarantee safety is that the creators are liable for what they build and generate.”
Pressed again by Vargas, he doubled down. “The one thing we should not do is give [AI labs] a blank check on liability,” he said.
“Because I believe that the best liability, or the best safety guard, is that they will be held responsible. And they are saying that ‘we would like to all slow down, but please give us a waiver on liability,’ which should not be done. And I would encourage everyone in this committee and in both houses not to consider that.”
The comments land days after Anthropic CEO Dario Amodei’s essay calling on companies to slow model development. President Donald Trump rejected easing the pace. OpenAI’s Sam Altman and xAI’s Elon Musk backed Amodei’s plan.
Amodei’s plan included a narrow federal waiver for certain safety conversations. OpenAI previously backed an Illinois bill to limit liability for AI-fueled harms, FedScoop notes. “Labs are asking for” liability relief is Bessent’s characterization.
After the July Hugging Face cyberattack, Bessent told Rep. Ritchie Torres the administration huddled with AI labs and the financial sector. He applauded the “largest banks” for “very good cybersecurity resilience.” Treasury’s work with big financial institutions has been formalized through Gold Eagle, a clearinghouse it oversees with the Cybersecurity and Infrastructure Security Agency for vulnerability scanning, validation, and patch distribution.
“Our mandate is with the financial sector,” Bessent told Rep. Josh Gottheimer. “And then we have attempted to migrate that knowledge and the learning from the financial sector into other sectors through the Gold Eagle forum.”
Bessent also called for more open-source models built in the United States to “push back against China.” “We can’t let these large labs have regulatory capture because that will stop innovation.”
Eric Revell at Fox Business prints the same open-source push and the China line: Chinese models “distill from the U.S. models, which is a polite scientific word for steal.” Distillation means copying capability from a closed model into another model.
The more the U.S. builds its own open models, Bessent said, the less people will use Chinese ones. He added that China’s distillation has left “many of the Chinese models [that] think they’re Mythos, they think they’re Claude.”
Fox notes a July joint letter against sweeping open-model restrictions signed by Nvidia, Microsoft, Meta, Dell, Palantir, Hugging Face, Mozilla, Mistral, and others.
While frontier CEOs argue about pacing and evaluators, the Treasury secretary’s safety answer to the House was simpler: keep the builders on the hook for what they ship, and don’t let a few closed labs write the rules.
No blank check. More U.S. open models. That is the testimony.