The U.S. Commerce Department lifted export controls on Anthropic's Fable 5 and Mythos 5 models, less than three weeks after imposing them on June 12 following a report from Amazon NASDAQ:AMZN researchers describing a method to bypass Fable 5's safeguards. Anthropic had suspended both models for all users. Now, Fable 5 will be available globally from Wednesday, with Mythos 5 access partially restored on June 26 to select U.S. organizations and is expanding to the broader Glasswing partner network.
Anthropic said it has deployed a new safety classifier that blocks the reported jailbreak technique in over 99% of cases, with blocked requests routed to its Opus 4.8 model. The company acknowledged the new classifier will flag some benign requests as a tradeoff.
Anthropic is also launching a jailbreak severity framework developed with Amazon, Microsoft NASDAQ:MSFT, Alphabet's (GOOGL) Google, and other Glasswing partners, designed to give AI developers and governments a common standard for assessing and responding to jailbreaks. OpenAI similarly delayed a full public rollout of GPT-5.6 at the U.S. government's request.