Anthropic ‘abruptly’ disables Fable 5, Mythos 5 to comply with US govt directive; says ‘this is a misunderstanding’
Anthropic has abruptly disabled global access to its Fable 5 and Mythos 5 artificial intelligence models after the US government issued an export control directive
Anthropic has abruptly disabled global access to its Fable 5 and Mythos 5 artificial intelligence models after the US government issued an export control directive on national security grounds, ordering the company to block all foreign nationals from using the systems, including Anthropic's own employees who do not hold US citizenship. Anthropic received the instruction at 5:21 p.m. ET on Friday, requiring it to suspend all access to both models by "any foreign national, whether inside or outside the United States, including foreign national Anthropic employees." What the US Government Ordered Anthropic to Do The directive, received on Friday at 5:21 p.m. ET, instructed Anthropic to suspend all access to the models "by any foreign national, whether inside or outside the United States, including foreign national Anthropic employees." To ensure full compliance, Anthropic extended the restriction to its entire customer base. In a statement, the company confirmed that access to all other Anthropic models would remain unaffected, and that it was working urgently to resolve the situation. Fable 5 and Mythos 5: What Was Pulled Offline The two models had only recently been unveiled, with Anthropic positioning them as state-of-the-art across a range of industry benchmarks. Fable 5 was particularly notable as the first time the company had released such a capable model for broad public use, underpinned by what Anthropic described as robust safeguards against misuse in high-risk domains, including cybersecurity.
The releases built upon Claude Mythos Preview, an advanced model that attracted considerable attention from both Wall Street and government officials when it was unveiled in April for its sophisticated cybersecurity capabilities. Anthropic had opted not to make Mythos Preview generally available, instead limiting its rollout to a small group of organisations as part of a cybersecurity initiative called Project Glasswing. Anthropic Says the Government Cited a Potential Jailbreak Although the government's letter did not specify the precise national security concern, Anthropic said its understanding was that officials believed a method for bypassing, or "jailbreaking," Fable 5 had been identified. The company, however, pushed back firmly on the gravity of the finding. In its full public statement, Anthropic said it had reviewed what it believed to be the evidence underpinning the directive, and described the capability displayed as widely available from other models already on the market. The company said the potential jailbreak involved asking the model to read a specific codebase and fix 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 a disclosure of a concerning non-universal potential jailbreak that led to a harmful result. The potential jailbreaks that have been disclosed to us are either entirely benign responses or are minor findings that provide no Mythos-specific uplift," Anthropic said. Anthropic's Defence-in-Depth Strategy for Fable 5 In its statement, Anthropic outlined the safeguard philosophy it had applied to Fable 5 prior to launch, noting that the model had undergone thousands of hours of red-teaming by the US government, the UK AI Safety Institute, and multiple private third-party organisations.
"These tests showed that Fable's safeguards are substantially more effective than those of any previously deployed model," the company said. Anthropic acknowledged that perfect resistance to jailbreaking was not achievable for any model provider at present, and said it had therefore adopted a defence-in-depth approach. The aim, it explained, was to make jailbreaks either narrow in scope or prohibitively expensive to produce, combined with thorough monitoring to detect and address any successful attempts quickly. This was also the rationale, the company said, behind its requirement for 30-day data retention by Fable customers, a policy it acknowledged carried commercial costs but considered necessary for ongoing safety research. "We suspect that perfect jailbreak resistance is not currently possible for any model provider. Every safeguard used in the industry is vulnerable to non-universal jailbreaks," Anthropic stated. Anthropic Complies But Calls the Action a 'Misunderstanding' Despite complying with the directive, Anthropic was unambiguous in its disagreement with the government's decision. The company warned that applying this standard consistently across the industry would effectively halt all new frontier model deployments. "We are complying with the government's legal directive and are removing access to Fable 5 and Mythos 5 for all users. However, we disagree that the finding of a narrow potential jailbreak should be cause for recalling a commercial model deployed to hundreds of millions of people," the company said. "If this standard was applied across the industry, we believe it would essentially halt all new model deployments for all frontier model providers." Anthropic also took aim at what it called the lack of transparency and due process in the action.
"As we have stated publicly, we believe the government should have the ability to block unsafe deployments, as part of a statutory process that is transparent, fair, clear, and grounded in technical facts. This action does not adhere to those principles," it added. A Deepening Rift Between Anthropic and US Federal Authorities The incident represents another chapter in what has become a prolonged and public dispute between Anthropic and the US government. Earlier this year, after negotiations between the company and the Department of Defense broke down, the DOD designated Anthropic a supply chain risk, a label that has historically been applied to foreign adversaries. The designation requires defence contractors to certify they will not use Anthropic's Claude models in military work. Anthropic subsequently filed suit against the Trump administration to challenge the blacklisting, and that litigation remains ongoing. Anthropic closed its Friday statement with an apology to affected customers and a commitment to transparency.
The BriefWire