Treasury Secretary Scott Bessent dismissed calls from leading AI executives for federal regulation, saying in an interview with Axios co-founder Mike Allen published Saturday that companies alarmed by their own technology should slow down themselves. International Business Times and Mediaite reported his remarks, in which Bessent also said the U.S. cannot afford to lose its AI lead to China.

What Happened

Allen asked Bessent about AI leaders, including Anthropic CEO Dario Amodei and OpenAI CEO Sam Altman, who have called for slowing the development of the most advanced models and who, Allen said, were essentially pleading for regulation. Bessent replied that in that case they should slow down, and compared the situation to Hannibal Lecter asking to be stopped before killing again, according to Mediaite.

Bessent described alarmism without solutions from parts of the AI community as a failure of leadership. At the same time, he said the country has to be prepared for every scenario, and IBTimes said he endorsed the idea of "safe acceleration," a slightly more cautious framing than the administration has used before.

RelatedAnthropic Asks Appeals Court to Rehear Ruling Upholding Pentagon's Supply-Chain Risk Label

On China, Bessent said the U.S. cannot lose its AI lead. He confirmed plans for a bilateral U.S.-China communications channel focused on AI, intended to coordinate responses to major incidents. He also accused Chinese open-source models of stealing from American ones, citing instances in which Kimi, a model from Beijing-based Moonshot AI, identified itself as Anthropic's Claude.

Why It Matters

Bessent's comments are among the clearest statements yet that the administration does not intend to answer the labs' warnings with binding rules. Last week, six AI companies signed a voluntary White House accord built on internal controls and outside audits, with no regulator or penalties, and President Donald Trump said he prefers that the industry police itself. Trump has called fears about AI a hoax and pledged not to burden AI companies with new regulation, Mediaite noted.

The Hannibal Lecter remark also turns a common criticism of the labs into administration policy: that executives who describe their products as potentially dangerous are in a position to change their own behavior and should not need Washington to force them. The labs' own case for rules rests on a collective-action problem: a company that slows down on its own risks losing ground to rivals, foreign or domestic, that keep going.

Context

The incident channel Bessent confirmed traces to last month's summit between Trump and Chinese leader Xi Jinping, after which a White House fact sheet described a U.S.-China dialogue on what both sides now call super intelligence and a channel for warning each other about incidents. Bessent pushed for the channel in talks with Chinese officials before the summit.

His comments on Kimi echo a broader dispute over distillation, in which developers train a model on another model's outputs. OpenAI recently said users linked to Moonshot had tried to extract reasoning from its systems to train their own.

IBTimes noted that Bessent has been mentioned as a possible candidate for a new AI czar role in the administration.

What to Watch

Trump said at the accord signing that he would name someone to oversee it, and the administration is expected to set up a new AI leadership post. Whoever fills that role, and whether it carries any enforcement power beyond voluntary commitments, will show how far the administration is willing to go beyond Bessent's message that the labs should police themselves.