President Donald Trump told Axios he no longer considers AI company Anthropic a national security threat — a complete reversal from just a week ago. The about-face came after a face-to-face chat at the G7 summit lunch, where Anthropic CEO Dario Amodei was awkwardly seated at the far end of the table.
Tensions had been simmering over the severity of jailbreak security vulnerabilities in Anthropic's Claude Fable 5 and Claude Mythos 5 models. Talks broke down, and on June 12, the Commerce Department hit the company with export controls, blocking the models from leaving the country and requiring special approval for foreign engineers to access them. That landed Amodei in an uncomfortable spot at the G7 lunch.
After the crisis point, negotiations shifted to joint development of technical standards. The White House, Commerce Department, national cyber director, and Anthropic executives held multiple rounds of talks. Safety experts were dispatched to DC to build a framework for assessing the severity of large model flaws — and where government intervention should stop. Trump said Amodei quickly cooperated with the demands.
The policy reversal clears a major obstacle for Anthropic's IPO and expansion. As of June 2026, the company had $47 billion in annualized recurring revenue, a $965 billion valuation, and had already filed for a public offering. Had the security threat label stuck, the regulatory uncertainty could have been a dealbreaker.
On broader tech regulation, Trump reiterated that he does not want to hobble the domestic AI industry by shutting down or seizing companies — his top priority is staying ahead of China. But he did not rule out using emergency executive powers, including the Defense Production Act. "I'm not sure yet if we need to pull that lever," Trump said, "but I will if I have to."