The company said Fable 5 and Mythos 5 were pulled worldwide after federal officials barred foreign access.
SAN FRANCISCO, CA — Anthropic disabled two of its most advanced artificial intelligence models worldwide after the Trump administration ordered the company to block access by foreign nationals, the company said Friday.
The order marks one of the strongest U.S. moves yet to control access to frontier AI systems, reaching beyond chips and data centers to the models themselves. Anthropic said the directive covered Fable 5 and Mythos 5, including access by foreign national employees inside the United States. The company said all other Anthropic models remained available.
Anthropic said it received an export control directive from the U.S. government citing national security authorities. The company said the order required it to suspend all access to Fable 5 and Mythos 5 by any foreign national, whether inside or outside the United States. Because Anthropic could not ensure compliance while keeping the tools online for some users, it said the practical result was a full shutdown for customers. “The net effect of this order is that we must abruptly disable Fable 5 and Mythos 5 for all our customers to ensure compliance,” the company said.
The directive came days after Anthropic announced Fable 5 and Mythos 5, describing Fable 5 as a model prepared for broad use and Mythos 5 as a more powerful system with tighter limits. Reports said officials were concerned the models could be used to find software flaws or that safeguards could be bypassed. Anthropic has said the government did not publicly provide detailed technical evidence for the restriction. The company also said it had worked with officials before launch and was seeking a path to restore access.
The shutdown quickly spread beyond Anthropic’s U.S. customer base because the order covered foreign nationals rather than only foreign countries. That meant the restriction could apply to non-U.S. citizens working in the United States, including some Anthropic employees. The rule also affected customers in allied countries, adding pressure on companies and governments that rely on U.S. AI providers. Canadian Prime Minister Mark Carney said the episode showed the risks of depending too heavily on a small number of American technology suppliers.
The order landed amid months of tension between Anthropic and the Trump administration over national security, military use and federal access to AI tools. Earlier disputes centered on how Anthropic’s safety rules applied to government work, including defense and intelligence uses. The new order moved the fight into export controls, a legal tool more often associated with semiconductors, military goods and other sensitive technology. AI companies and policy specialists said the case could set a precedent for how Washington handles powerful software that can be copied and used across borders.
Federal officials have not released a full public technical explanation of the decision. Reports cited concerns about a possible jailbreak and the models’ ability to help identify cybersecurity vulnerabilities. Anthropic has disputed the idea that the models had a broad, universal bypass and said other frontier systems can perform similar security tasks. The company said the affected models were built with safeguards and routing systems meant to limit dangerous use. It said those claims remain under discussion with the government.
The immediate question is whether Anthropic can persuade U.S. officials that new safeguards, access controls or identity checks are enough to lift the restriction. The company said it aims to restore access as quickly as possible, but gave no public date for the models’ return. The Commerce Department’s export control process could require more technical review, new licensing rules or limits on who may use the systems.
As of Sunday, Fable 5 and Mythos 5 remained suspended, while other Anthropic models continued operating. The next milestone is whether federal officials publish more details on the national security basis for the order or approve a narrower compliance plan.
Author note: Last updated June 14, 2026.