Anthropic restored global access to Claude Fable 5 on July 1 after the U.S. Department of Commerce lifted export controls imposed June 12. The original ban barred all foreign nationals—including Anthropic's own non-U.S. staff—from accessing Fable 5 and the more restricted Mythos 5 following Amazon researchers' discovery of a jailbreak technique that could prompt the model to identify software vulnerabilities and write exploit code.
To resolve the standoff, Anthropic deployed a single safety classifier that blocks the reported jailbreak in 99%+ of cases and reroutes flagged requests to the weaker Opus 4.8 model. The Commerce Department reviewed the safeguards over two weeks alongside Anthropic and Amazon. Mythos 5 returned to ~100 U.S. organizations on June 26; Fable 5 now ships on Claude.ai, Claude Code, and Claude Platform, with AWS/Google Cloud/Microsoft Foundry access to follow.
Commerce Secretary Howard Lutnick framed the reversal as strengthening U.S. AI leadership rather than penalizing Anthropic. The company committed to detecting and disclosing security risks, coordinating on future releases with the government, and reporting malicious activity. Anthropic also opened a HackerOne bug bounty for Fable 5 jailbreaks and promised designated federal agencies early access to frontier models pre-launch—a significant surveillance/coordination pact.
The 18-day ban briefly handed Chinese models like Z.ai's GLM-5.2 top benchmark positions while Fable was offline. This outcome signals a maturing government-AI company dynamic: export controls are now negotiable via technical remediation and information-sharing rather than outright bans, a critical precedent as frontier models face dual-use scrutiny.