Anthropic shut down access to its Claude Fable 5 and Claude Mythos 5 models following an emergency export control order from the United States government. The directive prohibits any foreign national from using the systems, including Anthropic staff who are not US citizens, and applies regardless of location. The order stems from security concerns over an alleged jailbreak method capable of bypassing protections built into Fable 5, though Anthropic disputes the rationale, stating the reported vulnerabilities are limited and previously known. The decision has reignited debate over the risks posed by advanced AI models and the appropriate regulatory response to potential security threats.
US Export Control Order Prohibits Foreign National Access to Claude Models
The company said the breadth of the order effectively required it to disable both models for all users. Anthropic noted that the restriction affects only Fable 5 and Mythos 5, with the remainder of its AI portfolio continuing to operate normally. Officials have not publicly outlined the specific national security risks behind the decision.
Anthropic Disputes Security Rationale Behind Model Shutdown
Anthropic said it understands the concern relates to an alleged jailbreak method capable of bypassing protections built into Fable 5. Following an internal review, the company said the demonstrated technique exposed only a small number of previously known vulnerabilities. According to Anthropic, the vulnerabilities identified through the technique were relatively simple and could also be discovered by other publicly available AI models without circumventing safeguards. The company stated that the evidence presented by authorities concerns a limited jailbreak scenario involving analysis of a specific codebase and the identification of software flaws. It also said similar capabilities are already available through competing models, including OpenAI's GPT-5.5.
Industry Implications of Recalling Commercially Deployed AI Models
Although Anthropic is complying with the directive, it maintains that recalling a commercially deployed model over a narrow potential jailbreak is unwarranted. The company warned that adopting such an approach across the industry could impede future frontier AI releases and said it is working to restore customer access.
FAQ
What did the US government order Anthropic to do?
The United States government issued an emergency export control order requiring Anthropic to shut down access to its Claude Fable 5 and Claude Mythos 5 models. The directive prohibits any foreign national from using the systems, including Anthropic staff who are not US citizens, and applies regardless of location.
Why does Anthropic dispute the security concerns behind the shutdown?
Anthropic disputes the security rationale, stating that an internal review found the alleged jailbreak method exposed only a small number of previously known vulnerabilities. The company said the vulnerabilities identified were relatively simple and could also be discovered by other publicly available AI models without circumventing safeguards, including OpenAI's GPT-5.5.
How does Anthropic view the impact of this decision on the AI industry?
Anthropic warned that recalling a commercially deployed model over a narrow potential jailbreak is unwarranted and that adopting such an approach across the industry could impede future frontier AI releases. The company is working to restore customer access.
