The U.S. government has ordered Anthropic to suspend foreign access to its new Fable 5 and Mythos 5 AI models, citing national security risks from a potential "jailbreak." To comply, Anthropic disabled the models for all users. The company publicly disagrees with the directive, calling it a misunderstanding and warning that the standard would halt all new AI deployments if applied industry-wide.
US export control directive
- ▪The directive applies to all foreign nationals, whether inside or outside the U.S., including Anthropic's own foreign national employees
- ▪Dean Ball, a former White House official, said the order suggests users should expect to have to prove their citizenship to use Anthropic's latest models
- ▪The order was issued by the Commerce Department and sent in a letter from Commerce Secretary Howard Lutnick
- ▪The U.S. government issued an export control directive ordering Anthropic to suspend foreign national access to its Fable 5 and Mythos 5 AI models, citing national security
Fable 5 model suspension
- ▪Fable 5 was a public-facing model with strong safeguards, while Mythos 5 was a less-restricted version shared with trusted partners for cybersecurity work
- ▪To ensure compliance, Anthropic "abruptly" disabled Fable 5 and Mythos 5 for all customers, not just foreign nationals
- ▪Access to all other Anthropic models is not affected by the directive
Government jailbreak concerns
- ▪The government's concern stems from a "jailbreak" method that could allow Fable 5 to be used to identify software vulnerabilities
- ▪Anthropic stated the government has only provided "verbal evidence" of a "narrow, non-universal jailbreak."
- ▪Anthropic claims the jailbreak's capability is widely available from other models, including OpenAI’s GPT-5.5, without needing a bypass
Defense-in-depth safeguard strategy
- ▪Anthropic's "defense in depth" strategy aims to make jailbreaks narrow or expensive to produce, combined with monitoring to detect attacks
- ▪Before launch, Anthropic, the US government, and third parties "red-teamed" Fable's safeguards for thousands of hours
- ▪No testers have yet found a "universal jailbreak" for Fable 5 that can broadly bypass its safeguards
Anthropic compliance disagreement
- ▪While complying with the order, Anthropic publicly disagreed with it, calling the situation a "misunderstanding."
- ▪Anthropic argued the government's action was not transparent, fair, or grounded in technical facts
Industry deployment standards
- ▪The Trump administration previously put Anthropic on a Pentagon supply chain blacklist after a dispute over military use of its AI
- ▪Anthropic warned that if this standard were applied industry-wide, it would "essentially halt all new model deployments" for frontier AI providers
- ▪The action is an escalation of U.S. policy, which has previously focused on controlling AI chips and tools rather than the models themselves
Story comments
Loading comments…