WASHINGTON, June 13 (Nationwide Times) — Anthropic said it was disabling access to its newest advanced AI models, Claude Fable 5 and Mythos 5, after a U.S. government directive aimed at blocking access by foreign nationals, according to the company’s statement as reported by the BBC and other accounts.
The action changes the status of the products at the center of the dispute: users who had access to the models were to lose it while Anthropic sought to comply with the government order. The public record available so far does not include the full directive, the full company statement, technical model cards or an independent test of the alleged security problem.
The clearest account of the government action comes from The Wall Street Journal, which reported that Commerce Secretary Howard Lutnick sent Anthropic CEO Dario Amodei a letter on June 13 stating that Fable 5 and Mythos 5 were subject to export limits. The Journal, citing the letter and an administration official, reported that the ruling barred use by clients outside the United States and by foreign nationals inside the United States.
Anthropic’s own explanation, as described by the BBC, said national-security officials had not identified a specific problem in the models. The company said its understanding was that the government believed there was a technique to bypass, or “jailbreak,” Fable 5.
Anthropic also said it reviewed a demonstration of that alleged bypass method and found that it exposed only a limited number of previously known, minor vulnerabilities, according to the BBC’s account of the company statement. No source in the supplied record provides independent technical verification of the alleged jailbreak, the exact vulnerabilities or whether disabling access for all users was the only technical way to comply.
That gap matters because the government’s stated concern appears to be about who could use the models, not only what the models could do. The Journal’s account frames the June 13 action as an export-limit decision. The BBC and other accounts describe it more generally as a government directive or order. The specific legal mechanism has not been made public in the materials reviewed for this story.
The shipping status of the models is also partly documented and partly unclear. The BBC reported, based on Anthropic’s explanation, that Fable 5 had been released privately in April for preview and vulnerability testing and was launched publicly only days before the suspension. The supplied record does not include official release notes, exact launch dates or detailed capability descriptions for Fable 5 or Mythos 5.
The June suspension follows months of conflict in Washington between Anthropic and the Trump administration over federal and military use of the company’s Claude models.
In February, Anthropic said it refused Pentagon demands to remove safety restrictions for two categories of use: mass domestic surveillance and fully autonomous weapons capable of lethal action without human involvement. Those limits were described in public statements by Anthropic and Amodei, according to accounts by the BBC, The New York Times and other national outlets.
President Donald Trump later said federal agencies should stop using Anthropic technology, according to the BBC and The New York Times, which cited Trump’s public post and Anthropic’s response. Defense Secretary Pete Hegseth also publicly criticized the company over access to its models for lawful military use, according to NBC News and other accounts.
The Pentagon or Defense Department then designated Anthropic a “supply chain risk,” according to Anthropic statements and Defense Department accounts described by NBC News, the BBC and CNBC. Anthropic said that label would bar defense contractors from using Anthropic tools on defense-related work.
Anthropic said it would challenge the designation in court. The BBC later reported that a judge ruled the Pentagon order could not be enforced while the lawsuit proceeds. The available record does not name the court, judge, case caption or docket number.
The dispute also affected federal operations in Washington. Reuters reported in March that the State Department, Treasury Department and Department of Health and Human Services had begun moving away from Anthropic products. Reuters cited a State Department memo, an email from State Department spokesperson Tommy Pigott and a public post by Treasury Secretary Scott Bessent.
The State Department memo said its internal chatbot, StateChat, would use GPT4.1 from OpenAI during the transition, according to Reuters. Reuters also reported that federal housing agencies, including the Federal Housing Finance Agency, Fannie Mae and Freddie Mac, would stop using Anthropic products.
Anthropic’s role in federal AI systems had grown before the dispute. The BBC reported that Anthropic had been used by the U.S. government and military since 2024. Multiple national outlets reported that Anthropic had been the only AI model approved for classified military systems until early 2026, though the approval document is not included in the supplied record.
The records now public show two related but distinct federal actions: a defense-procurement dispute over military access to Anthropic’s models, and a later Commerce-linked restriction that led Anthropic to disable Fable 5 and Mythos 5. What the records do not yet show is the full government order, the full export-control basis, the full court record or independent evidence proving the alleged Fable 5 jailbreak method.
