US Government Export Order Forces Anthropic to Shut Down Fable 5 and Mythos 5 Worldwide

The US government has reportedly ordered Anthropic to restrict access to its most advanced AI models, Fable 5 and Mythos 5, to foreign nationals. In response, Anthropic has suspended both models globally, affecting users worldwide.

According to the company, it received the directive on June 12 at 5:21 PM ET under national security authority provisions. The order requires that access to Fable 5 and Mythos 5 be blocked not only for users outside the United States but also for foreign nationals inside the US, including Anthropic’s own employees.

Because of the broad scope of the restriction, Anthropic says it had no option but to disable both models for all customers. Other models, including Claude Opus 4.8, remain available and are not impacted.

Fable 5 had only recently been rolled out on June 9 as a temporary free release for Pro, Max, and Enterprise users, scheduled to continue through June 22. That rollout has now been abruptly cut short, leaving users unable to access the model just days after its release.

Fable 5 is described as a safety-filtered version of Mythos 5. While both are based on the same underlying system, Fable includes additional safeguards that block or redirect sensitive queries in cybersecurity, biology, and chemistry domains. Mythos 5, by contrast, is a less restricted version available only to vetted government and research partners.

Anthropic’s developer notice explained that new sessions default to alternative models such as Claude Opus 4.8, while any active Fable 5 sessions will fail with errors. Developers using the API have been instructed to migrate away from the affected models.

The decision has sparked debate. A UK AI and online safety minister highlighted the disruption affecting both US and UK users and framed it in the broader context of technological sovereignty and AI infrastructure investment.

Anthropic, however, disputes the severity of the government’s concerns. The company says the directive was based on a claimed “narrow jailbreak,” which it describes as a limited and non-general exploit involving prompting the model to analyze code and identify vulnerabilities. Anthropic argues that similar capabilities are already widely known and present in other AI systems.

The company also stated that it only received verbal evidence of the issue and believes the reported vulnerability does not justify the removal of a widely deployed commercial model.

Anthropic further warned that applying such strict standards across the industry could significantly slow or even halt the deployment of new frontier AI systems.

Despite its disagreement, the company confirmed it is complying with the directive while working to restore access and plans to release additional details soon.