Claude Fable 5 Returns After a 19-Day Government-Ordered Shutdown

    The Commerce Department lifted the first export controls ever applied to a deployed AI model, and Anthropic restored Fable 5 worldwide on July 1 with stronger safeguards and a proposed industry standard for judging jailbreaks.

    Claude Fable 5 Returns After a 19-Day Government-Ordered Shutdown The Commerce Department lifted the first export controls ever applied to a deployed AI model, and Anthropic restored Fable 5 worldwide on July 1 with stronger safeguards and a proposed industry standard for judging jailbreaks. Aaron Rafferty July 02, 2026 Key Takeaways The US government lifted export controls on Anthropic's Claude Fable 5 and Mythos 5 on June 30, ending a 19-day suspension that had cut off both models for every user worldwide. The June 12 directive followed a report from Amazon researchers describing a way to bypass one of Fable 5's cybersecurity safeguards. Anthropic says its testing showed less capable models could match the reported behavior. Anthropic returned the model with a retrained safety classifier that blocks the reported technique in over 99 percent of cases, and is drafting a jailbreak severity framework with Amazon, Microsoft, and Google. Anthropic restored global access to Claude Fable 5 on July 1 after the Department of Commerce lifted the export controls that had kept the company's most capable public model offline for 19 days. It is the first time the US government has used export authorities to pull a widely deployed commercial AI model from the market, and the reversal came just as quickly. The June 12 directive, citing national security, barred any foreign national from accessing Fable 5 and Mythos 5. Because Anthropic had no reliable way to verify nationality in real time, it shut both models off for everyone. The trigger was a report in which Amazon researchers found a method of bypassing Fable 5's safeguards to identify software vulnerabilities. Anthropic disputed the severity, saying models including GPT-5.5 could produce the same results, and rese

    Loading full article…