U.S. Export Controls Suspend Access to Anthropic's Mythos 5 and Fable 5 Models
The U.S. Commerce Department issued an export-control directive requiring Anthropic to block foreign-national use of its Mythos 5 and Fable 5 AI models, prompting negotiations between the company and the Trump administration.
Anthropic received a U.S. Commerce Department export-control directive on Friday ordering the company to suspend access to its Mythos 5 and Fable 5 artificial-intelligence models for any foreign national, including its own employees. The directive gave the firm a short window to comply, after which broader export restrictions could be imposed.
Anthropic said the order stemmed from a government concern that a method for bypassing Fable 5’s safeguards had been demonstrated. The company described the reported “jailbreak” as a narrow, non-universal technique that it had shared with the authorities, and noted that similar capabilities exist in other large language models, including OpenAI’s GPT-5.5.
According to a source involved in the talks, the administration contacted Anthropic early Friday and set a 90-minute deadline for shutdown. Anthropic’s chief executive, Dario Amodei, and senior staff met with officials from the Treasury, Commerce and the National Cybersecurity Office later that day to discuss the directive.
The concern was reportedly triggered by indications that a China-linked entity may have accessed the technology, prompting Anthropic to revoke that party’s access to the earlier Mythos Preview model. Some reports also cite Amazon researchers who flagged potential vulnerabilities in Fable 5 during internal red-team testing.
Industry leaders have responded with a public letter urging the administration to lift the restrictions, arguing that the models are not uniquely capable of exploiting vulnerabilities and that overly aggressive safeguards can hinder legitimate security work. Alex Stamos, chief product officer at Corridor, said policymakers may underestimate the speed at which foreign competitors can close the gap with U.S. AI developments.
The dispute comes as other major AI firms—OpenAI, Google and Microsoft—have released comparable models, raising questions about whether similar export controls could later be applied to them. The negotiations remain ongoing, with no final resolution reported as of Monday.
