The U.S. government on Friday ordered Anthropic to suspend foreign access to its Fable 5 and Mythos 5 artificial intelligence models, citing national security concerns tied to a reported jailbreak method. The company said it disabled the models the same evening to comply.
KEY FACTS
- Directive Commerce Secretary Howard Lutnick told Anthropic to stop foreign access to the two models.
- Scope The export controls bar use by foreign nationals inside or outside the United States.
- Company response Anthropic disabled the models and said other AI systems were not affected.
- Dispute Anthropic said the concern involved a narrow jailbreak and was not a universal bypass.
Fable 5 and Mythos 5 had been released earlier in the week, with the company describing them as its most capable systems. Mythos 5 was also available to members of Project Glasswing, a program for selected cybersecurity companies to use the model for finding and fixing security flaws.
The letter from the Commerce Department did not spell out the national security concern. In a Friday night company disclosure, Anthropic said it understood the government had learned of a technique for bypassing safety guardrails that involved prompting the model to read a specific codebase and identify software flaws.
The company said it reviewed a report it believes underpinned the directive and found that the capabilities shown were already available in other public models, including OpenAI’s GPT-5.5. It also said the same tools are used routinely by cybersecurity professionals for defensive work.
Anthropic said no testers had found a universal jailbreak that broadly defeated the safeguards. It added that perfect jailbreak resistance is not achievable and said the models were built with a defense-in-depth approach that combined narrow jailbreak resistance with active monitoring.
The move follows a wider dispute between Anthropic and the Trump administration. The company said it is working to restore access as soon as possible, while researchers and analysts questioned the breadth of the export controls and the inclusion of foreign nationals inside the United States.
WHY IT MATTERS
The action could set a precedent for restricting access to advanced AI models based on safety concerns rather than only their output or deployment. It also raises questions for companies that build and test frontier models with international staff and users.