Anthropic ‘abruptly’ disables Fable 5, Mythos 5 to comply with US govt directive; says ‘this is a misunderstanding’

Anthropic abruptly disabled global access to its Fable 5 and Mythos 5 AI models after the U.S. government issued an export control directive on national security grounds, ordering Anthropic to block all foreign nationals, including its own employees who do not hold U.S. citizenship, from using the systems.

Anthropic received the order at 5:21 PM ET on Friday, requesting that it suspend all access to both models by “any foreign national, whether within or outside the United States, including foreign national Anthropic employees.”

To read Anthropic’s Full Disclosure

What Did the US Government Order Antropik to Do?

The directive, received at 5:21 p.m. ET on Friday, instructed Anthropic to suspend all access to the models by “any foreign national, whether within or outside the United States, including foreign national Anthropic employees.” Anthropic has extended the restriction to its entire customer base to ensure full compliance.

In a statement, the company confirmed that access to all other Anthropic models will not be affected and that it is working urgently to resolve the situation.

Fable 5 and Mythos 5: Those Taken Offline

The two models were recently introduced, with Anthropic positioning them as cutting-edge based on various industry benchmarks. Fable 5 was particularly notable as it was the first time the company had released such a capable model for broad public use, backed by what Anthropic describes as robust protections against abuse in high-risk areas, including cybersecurity.

The versions were based on the Claude Mythos Preview, an advanced model that attracted significant attention from both Wall Street and government officials when it was introduced in April for its advanced cybersecurity capabilities. Anthropic chose not to make the Mythos Preview generally available, instead limiting its distribution to a small group of organizations as part of a cybersecurity initiative called Project Glasswing.

Anthropic Says Government Talks About a Potential Breakout

Although the government’s letter did not specify any specific national security concerns, Anthropic said officials believe a method to bypass or “jailbreak” Fable 5 has been identified. However, the company firmly backed down in response to the seriousness of the finding.

In its full public statement, Anthropic said it had reviewed the evidence believed to underpin the directive and identified that this demonstrated capability was widely available in other models currently on the market. The company said the potential jailbreak would require reading a specific code base from the model and fixing software flaws, and concluded that the same result could be achieved using publicly available tools, including OpenAI’s GPT-5.5.

“We have not even received disclosure of a worrisome, non-universal potential jailbreak that results in a harmful outcome. The potential jailbreaks disclosed to us are either completely benign responses or minor findings that do not provide a Mythos-specific improvement,” Anthropic said.

Anthropic’s Defense in Depth Strategy for Fable 5

In its statement, Anthropic outlined the protection philosophy it applied to Fable 5 before launch, noting that the model had been subjected to thousands of hours of red teaming by the US government, the UK AI Security Institute and numerous private third-party organizations.

“These tests demonstrated that Fable’s protections are significantly more effective than any previously used model,” the company said.

Anthropic acknowledged that perfect resistance to jailbreaking is currently not achievable for any model provider, so it said it was adopting a defense-in-depth approach. The aim, he explained, is to narrow the scope of jailbreaks or make them prohibitively expensive to produce, and combine them with comprehensive monitoring to quickly detect and address successful attempts. The company said this was also the rationale behind the 30-day data retention requirement for Fable customers, a policy it acknowledged carried commercial costs but felt was necessary for ongoing security research.

“We suspect that perfect jailbreak resistance is not possible for any model provider at this time. Every protection used in the industry is vulnerable to non-universal jailbreaks,” Anthropic said.

Anthropik Complies But Calls Action ‘Misunderstanding’

Despite complying with the directive, Anthropic was clear that it was against the government’s decision. The company warned that consistent application of this standard across the industry would effectively halt all new lead model deployments.

“We are complying with the government’s legal directive and removing all users’ access to Fable 5 and Mythos 5. However, we disagree that a narrow finding of potential jailbreak should be a reason for a recall of a commercial model applied to hundreds of millions of people,” the company said. he said. “If this standard were implemented industry-wide, we believe it would essentially halt all new model deployments for all lead model providers.”

Antropik also took aim at what he called a lack of transparency and due process in the action.

“As we have stated publicly, we believe that the government should have the ability to prevent unsafe deployments as part of a legal process that is transparent, fair, clear, and based on technical facts. This action does not comply with those principles.” he added.

A Deepening Separation Between Anthropic and U.S. Federal Officials

The incident represents another chapter in what has become a long-running and public dispute between Anthropic and the US government. Earlier this year, after talks between the company and the Department of Defense collapsed, the Department of Defense designated Anthropic as a supply chain risk, a label historically applied to foreign adversaries. The designation requires defense contractors to certify that they will not use Anthropic’s Claude models for military work. Antropik later sued the Trump administration to challenge the blacklisting, and that lawsuit is still ongoing.

Anthropic closed its statement on Friday with an apology to affected customers and a commitment to transparency.

“We apologize to our customers for this outage. We believe this was a misunderstanding and are working to restore access as quickly as possible,” the company said. the company said, adding that it plans to release more technical details within 24 hours.