Anthropic has suspended global access to its frontier AI models, Fable 5 and Mythos 5, effective immediately. The drastic measure follows a US government export control directive received today, citing national security authorities. The company stated that the order mandates suspension of access for all foreign nationals, a condition that necessitated disabling the models for every customer to ensure full compliance.
The directive landed at 5:21 PM Eastern Time. According to Anthropic, the government letter lacked specific details on the national security concern. However, the company’s understanding is that officials believe they have identified a method to bypass, or “jailbreak,” Fable 5’s safeguards. Access to all other Anthropic models remains unaffected.
Company Disputes Severity of the Alleged Jailbreak
Anthropic pushed back strongly against the action, characterizing the evidence presented by the government as a minor, non-universal jailbreak. The technique in question involves prompting the model to review a codebase and identify software flaws. The company said it reviewed the reported method and validated that the capability it demonstrates is already widely available from other publicly accessible models, 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 company stated in its announcement. “The potential jailbreaks that have been disclosed to us are either entirely benign responses or are minor findings that provide no Mythos-specific uplift.”
Defense in Depth Strategy Under Scrutiny
The company defended its safety protocols, reiterating a “defense in depth” strategy that combines strong safeguards with rigorous monitoring. Anthropic acknowledged that perfect jailbreak resistance is likely impossible for any model today and was transparent about this limitation during Fable 5’s launch. The strategy aims to make jailbreaks either narrow in scope or prohibitively expensive to develop.
Key points of Anthropic’s safety posture include:
- Thousands of hours of red-teaming were conducted with the US government, UK AISI, and third-party organizations before launch.
- Fable’s safeguards were found to be substantially more effective than any previously deployed model.
- No tester has yet discovered a “universal jailbreak” capable of broadly bypassing all safeguards.
- A controversial 30-day customer data retention policy was implemented specifically to enable research and mitigation of jailbreaks.
Warning of Industry-Wide Implications
Anthropic is complying with the legal order but warned that the precedent set by this decision could have a chilling effect. The company argued that if a narrow, non-universal jailbreak finding is sufficient grounds to recall a commercial model used by hundreds of millions, it would essentially halt all new deployments from every frontier model developer.
“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,” the statement read. “This action does not adhere to those principles.”
The company apologized for the disruption and said it believes the situation stems from a misunderstanding. Efforts are underway to restore access as quickly as possible, with a promise to share more technical details within the next 24 hours.
