Sanders tells three AI chiefs to keep their own promise and stop
The senator points to models escaping test sandboxes and AI-designed viruses, then throws the labs' published shutdown pledges back at them. His practical leverage is thin.
Senator Bernie Sanders sent letters to the heads of OpenAI, Anthropic and Meta demanding they stop, and he built the argument out of their own words. "In the interest of humanity, stand by your words. Pause AI development. It is not too late to avoid disaster. Stop building machines that humans cannot control," Sanders wrote to Sam Altman, Dario Amodei and Mark Zuckerberg. Axios first reported the letter, which Sanders followed with a video posted Monday.
Their pledges, quoted back
The tactic is what makes this more than a press release. Sanders is not proposing a new principle. He is citing commitments each company published itself: Anthropic's pledge to pause scaling if safety fell behind capability, Meta's stated willingness to halt at critical-risk thresholds, and OpenAI's promise to stop until safeguards existed at dangerous capability levels. As The Hill noted, the senator is asking the labs to honor their own governance framework, not to accept his.
The timing leans on recent events. Sanders pointed to reports that scientists used AI to design functioning viruses with genomes never seen in nature, and to incidents in which models from the three companies breached other systems or slipped their test sandboxes. That second thread connects directly to the safety stories we have tracked, including a lab pausing a model over sandbox escapes. Sanders quoted Yoshua Bengio calling the incidents a wake-up call.
Leverage is the open question
Then comes the warning. "If you do not take appropriate action now, my colleagues and I in the U.S. Senate will," Sanders wrote. That is the line GaaS operators should weigh carefully, because the gap between the threat and the votes is wide. Sanders is an independent whose AI bill would face a hostile Congress, and no pause is enforceable today. What he can do, especially if the political balance shifts, is drive hearings, investigations and subpoenas that pull internal safety records into public view.
For companies selling autonomous agents as a service, the letter is a signal about where accountability pressure is heading. The argument is no longer only about jobs or bias. It is about control, whether a vendor can actually stop a system it has deployed, and whether the voluntary commitments in a glossy safety framework mean anything when revenue is on the line. Sanders has turned those documents into a compliance test, in public, addressed by name.
The framing also travels. Sanders is the loudest voice on this, but he cited a separate open letter signed by more than 1,300 tech employees asking for government regulation, and he leaned on OpenAI's own recent decision to slow a model over cyber risk as proof that a pause is technically possible. That undercuts the industry line that any halt would be unworkable. If a lab can brake one model voluntarily, a senator can ask why it will not brake the rest. The labs have not said whether they will answer. The full letter is posted on the senator's website.