Washington: Anthropic announced on Friday that it is disabling its "Fable 5" and "Mythos 5" AI models for all users after receiving an export control directive from the US government, which cited national security concerns.
According to Anadolu Agency, the US government's directive mandates the suspension of access to Fable 5 and Mythos 5 by any foreign national, regardless of their location, including those employed by Anthropic. The company clarified that this restriction does not affect access to any of its other AI models.
Anthropic reported receiving the directive at 5:21 pm ET (2121GMT) on Friday, noting that the government's letter did not provide specific details regarding the security concerns.
The company indicated that the government suspects a potential "jailbreak" or bypass of Fable 5's safety controls. However, Anthropic described the issues as "minor" and acknowledged that similar vulnerabilities could be found in other publicly available AI models without any special bypass methods.
Anthropic emphasized that extensive testing of Fable 5's safety systems was conducted in collaboration with the US government, the UK's AI safety body, and other organizations prior to the model's release. Despite these efforts, no method to fully bypass the model's protections was discovered.
The company acknowledged that achieving perfect jailbreak resistance is unlikely for any AI model provider at present, as all AI safety systems possess some weaknesses and potential bypass methods may emerge over time.
Anthropic revealed that the government provided only verbal details of one limited bypass method, which involved "asking the model to read a specific codebase and fix any software flaws." The company reviewed a report believed to be behind the directive and found that similar capabilities are available in other models, including OpenAI's GPT-5.5, and are commonly used for system defense.
While Anthropic agreed to comply with the government's order, it expressed disagreement over the decision to withdraw a widely-used commercial AI model due to the limited issue identified. The company argued that applying this standard industry-wide would impede the deployment of new frontier models.
Anthropic attributed the situation to "a misunderstanding" and assured users that efforts are underway to restore access to the models promptly.