US lifts export controls on Anthropic’s frontier cybersecurity AI models

Anthropic temporarily lost global access to its Fable 5 model after U.S. export controls were applied following a reported 'jailbreak' technique; after reaching agreements with regulators the company implemented mitigations (a new safety classifier, expanded government pre-release testing, a bug bounty, and participation in shared security standards) and access was restored, while access to its Mythos 5 model remains restricted to vetted U.S. organizations. The episode highlights tensions between enabling defensive security capabilities and controlling frontier AI models, and has prompted industry and government actions to formalize evaluation and disclosure practices.