At 5:21 p.m. ET on Friday, Anthropic received a Commerce Department directive invoking national security export-control authority to bar foreign-national access to Fable 5 and Mythos 5, the company’s two most powerful models. Because the order applied to any foreign national “whether inside or outside the United States,” including Anthropic’s own foreign-national employees, the company concluded it had no practical option but to shut both models down globally. Other Anthropic models remain available.

It’s the first time Washington has used export-control authority to pull a deployed frontier AI system off the market.

The Commerce letter gave no specific account of the concern. Anthropic’s understanding, per its public statement, is that the government believes it has identified a method of jailbreaking Fable 5. After reviewing a demonstration, the company called the technique “previously known, minor vulnerabilities” that other publicly available models, including OpenAI’s GPT-5.5, can surface too.

The political mechanism is more revealing than the legal one. Fortune reported that Amazon researchers elicited restricted cyberattack information from a Mythos-class model and that Amazon chief executive Andy Jassy escalated the matter to senior administration officials on Thursday. White House AI adviser David Sacks said on X that Anthropic chief executive Dario Amodei was asked to patch the issue or withdraw the model, and that the export order followed his refusal. Senior Anthropic technical staff spent the weekend in Washington meeting White House officials. The company says it disagrees with the action but is working to restore access.

There’s structural background here too: litigation between Anthropic and the Trump administration over an earlier Defense Department supply-chain designation is still pending. A safety-forward lab built its brand on warning regulators about model risk. The regulators, prompted by a cloud partner, have now used that framing against it.

Sources