U.S. bars non-Americans from Anthropic AI models; company shuts both models to all users

U.S. Commerce Department sent Anthropic a letter on Friday afternoon that banned non-Americans, including the company’s own employees, from accessing Fable 5 and Mythos 5. The directive invoked an export control rule and cited an unspecified national security concern.

Anthropic responded by shutting down both models to all customers. The company said it believes the order relates to a reported guardrail bypass but noted that the letter provided no specific details. The letter has not been made public.

Anthropic said the lack of clarity left it no choice but to restrict access to remain in compliance. Katie Moussouris, a cybersecurity researcher, said Anthropic shared with her a private copy of a paper describing an alleged guardrail bypass in Fable 5. The Wall Street Journal reported that the paper’s authors are security researchers at Amazon.

Moussouris stated that the bypass described in the paper should never have triggered an export control. She said the difference between asking a model to review code for security issues and asking it to fix code is largely semantic, yet the government treated the latter as a national security matter. Moussouris criticized the directive as hasty and heavy-handed.

She said the behavior described in the paper cannot meaningfully be fixed without weakening the model’s defensive capabilities. Dozens of other security researchers have called on the Trump administration to revoke the order.

U.S. Network defenders as dangerous. TechCrunch reported that the administration has not confirmed why it invoked the export control. U.S. export rules on cybersecurity tools were written so broadly that they nearly restricted legitimate security research in the 2010s.