Bessent aimed his criticism at AI leaders who warn without offering solutions, in an Axios interview published on Saturday, October 3, 2026. U.S. Treasury Secretary Scott Bessent said: "This alarmism without solutions by some of the AI community, that's not leadership." In September, Anthropic's Dario Amodei, OpenAI's Sam Altman and xAI's Elon Musk had called for the industry to slow development of its most advanced models, and Bill Gates warned the technology could cause a billion deaths. Bessent's remarks are Washington's formal answer to that round of warnings.
The risk is the labs' own ledger
Bessent's core claim is about ownership: the people who should manage a risk are the people creating it. "It is the people at the labs who have to accept responsibility, and I agree with that," he said. Asked about executives who call for regulation, he was blunter: "Well, then they should slow down." In his framing, if a company truly believes its model might spin out of control, the first move is to adjust its own release pace — not to ask the government for rules, guardrails or a liability shield first.
That lines up with the White House's position this week: after meeting technology executives, Donald Trump rejected new federal regulation and instead backed a non-binding arrangement to assess system safety through external audits, stressing that self-regulation is very important. Bessent added that he believes the labs "have come around to that way of thinking, too." A Treasury secretary's words carry weight here because Treasury sits at the center of financial stability, export controls and investment screening; with its chief talking this way, the bar for hard federal AI safety legislation in the near term just rose again.
Watch for plans now, not statements
The backdrop is a run of months in which warnings stopped being hypothetical: models acting on their own during evaluations, safety staff quitting publicly and criticizing company culture — the resignation of OpenAI's safety transparency lead is the latest example. Bessent did not deny that the risks exist. He denied the value of merely announcing them.
For the labs, the practical consequence is clear: before the next warning letter, attach what you have actually done — which product line you slowed, which audit you added, which external check you will accept. Without that, Washington will hear warnings as bargaining chips. For regulators and investors elsewhere, the signal is just as clear: the federal answer for now is self-regulation plus external audits, and binding rules are more likely to grow first out of state law, courts and sector regulators. When the next hearings and audit arrangements land, whoever puts a plan on the table first will be the one taken seriously.
