Anthropic’s most powerful AI model goes dark worldwide – just 3 days after launch

Anthropic released Claude Fable 5 and Mythos 5 on June 9 – its most capable
public models yet, with Fable 5 adding safety guardrails on cybersecurity,
biology, and chemistry queries. Three days later, an export control directive
ordered all foreign nationals blocked from both models, including Anthropic’s
own staff. Unable to comply selectively, Anthropic disabled both globally for
everyone. The cited reason: a jailbreak of Fable 5’s safeguards – which Anthropic
calls minor and “a misunderstanding,” and says it’s working to restore access.
Other Claude models remain unaffected.

This is reportedly the first time a frontier AI model has been forced offline globally
post-launch. Enterprises relying on Fable 5 or Mythos 5 should expect sudden
unavailability risk for frontier models tied to national security disputes, and build
fallback plans (e.g., Opus 4.8) into any agentic workflows using models.