Claude Fable 5 accessible again after export controls lifted

Anthropic restores Claude Fable 5 access on July 1 after the US lifts export controls, deploying a new classifier to block the security bypass method.

Suspended since mid-June, Claude Fable 5 will once again be accessible worldwide on July 1st, on Claude Platform, Claude.ai, Claude Code, and Claude Cowork. This return follows the lifting, on June 30th, of export controls that the U.S. government had imposed on this model, as well as on Claude Mythos 5. The restrictions targeted foreign nationals. Due to the inability to verify nationality in real-time, Anthropic had cut off access to both models for all users.

The directive originated from an Amazon researchers' report describing a method to bypass Fable 5's safeguards, even producing, in one instance, code exploiting a software vulnerability. A new classifier now blocks the reported technique, according to the company, in over 99% of cases. Some code and debugging requests will temporarily switch to Opus 4.8 while false positives are reduced.

With Amazon, Microsoft, Google, and other partners in the Glasswing program, Anthropic is also beginning to draft a common framework for measuring the severity of jailbreaks and calibrating developers' responses, with an open invitation to other model providers. Its collaboration with the U.S. government is simultaneously expanding to include access to models before their release and information sharing on circumventions. Access to Claude Mythos 5, meanwhile, has been restored for a set of U.S. organizations.