Anthropic disables access to Mythos models after US government order
Anthropic halted availability of its Fable 5 and Mythos 5 AI models following a Commerce Department directive that bars foreign nationals from using the technology, citing national-security concerns.
Anthropic PBC said it has shut down access to its latest artificial-intelligence models, Fable 5 and Mythos 5, after receiving a Commerce Department letter ordering the company to block all foreign users, whether inside or outside the United States. The directive, described by the firm as unprecedented, was issued on national-security grounds.
The company indicated the order came after researchers at Amazon identified a potential “jailbreak” that could allow the models to bypass safety guardrails, particularly in cybersecurity tasks. Anthropic’s statement said it disagreed that a narrow vulnerability should trigger a recall of a commercial model used by millions, warning that such a standard could halt new deployments across the industry.
“The Admin’s hope now is that Anthropic remediates the safety issue, the export control is lifted, and Fable goes back into general release,” wrote David Sacks, a former Trump AI adviser, on the social platform X.
“For anyone who was naive and perhaps hoping that this leverage wouldn’t be exerted, it’s a massive wake-up call,” said Aidan Gomez, co-founder of Cohere, in an interview.
Amazon declined to provide details of its discussions with U.S. officials, noting that government consultations on security risks are routine. The move marks the first time the U.S. government has imposed such broad restrictions on frontier AI software, raising questions about constitutional and commercial implications.
Industry leaders, including executives from Nvidia and OpenAI, have previously urged the government to promote global adoption of American AI rather than limit it. The European Commission said it is reviewing Anthropic’s remarks and continuing talks with allies about the cybersecurity risks of powerful AI models.
Anthropic, which is preparing for a potential initial public offering, had earlier called for a framework allowing governments and developers to pause AI work deemed dangerous. The current restriction adds pressure on the company to address the identified vulnerability and could set a precedent for other AI developers.
