BANNED AND BACK: ANTHROPIC’S MOST POWERFUL AI MODELS RETURN AFTER US LIFTS EXPORT CONTROLS Fable 5 and Mythos 5 are back online globally after a 19-day shutdown triggered when Amazon researchers discovered a jailbreak that made the models identify software vulnerabilities and generate exploit code. The Commerce Department banned them on June 12. On June 30 it lifted the controls. On July 1, Anthropic flipped the switch. The trigger was ugly. Researchers at Amazon found a way to bypass Fable 5’s safety guardrails and get it to produce working exploit code for software vulnerabilities. The June 12 order was so broad it required Anthropic to cut off access for any foreign national anywhere in the world, including its own non-citizen employees. With no reliable way to verify every user’s nationality in real time, Anthropic shut both models down for everyone. Anthropic says it has now blocked the jailbreak technique in more than 99 percent of attempts. Any blocked request gets rerouted to Opus 4.8 with a notification to the user. Mythos 5, which shares the underlying architecture but carries fewer safety restrictions, has been restored only to a select group of US organizations. The rest of the world gets Fable 5 back. What the 19-day shutdown proved, for anyone who needed the lesson, is that AI has become infrastructure. When it goes down, the damage is immediate, measurable, and hard to explain to customers. Keywords: Anthropic Fable 5 restored, Claude Mythos export controls, AI export ban lifted, Anthropic jailbreak