Overview
- The Commerce Department revoked the export‑control requirement for Claude Mythos 5 and Claude Fable 5 on Wednesday, and Anthropic has begun restoring access to both models.
- Anthropic is redeploying Claude Fable 5 worldwide with new classifiers that block queries that could be used for cyberattacks, and it will restore Mythos 5 only to a limited set of U.S. organizations approved through the Glasswing program.
- The decision followed a report from a trusted partner, identified by several outlets as Amazon Web Services, that found a method to bypass Fable’s safeguards; Anthropic implemented technical mitigations and additional security measures in response.
- As part of the deal, the company agreed to proactive risk detection, to report suspected malicious use to U.S. authorities, and to work with the government on launch protocols and standards, with officials reserving the right to reimpose controls if commitments lapse.
- The episode sets an early precedent for the Trump administration’s formal review of 'frontier' AI, raises industry concerns that tight controls could slow innovation, and comes as rival models from China and Japan increase pressure on U.S. policy makers.