White House Lifts Export Ban on Anthropic's Fable 5, Mythos 5

A US export control fight over a frontier AI model, announced and then reversed inside three weeks, is the kind of story that would have sounded absurd to anyone reading policy briefings a year ago, which is why the Anthropic-Commerce Department resolution on June 30 is worth reading carefully. According to Politico and corroborating reporting, the Commerce Department lifted the export controls it had imposed on Claude Fable 5 and Mythos 5, about two and a half weeks after they were issued.

The original June 12 order told Anthropic to cut off both models for any foreign national, inside or outside the United States, including its own non-citizen staff. Researchers at Amazon, Anthropic's largest investor, had discovered a 'fix this code' jailbreak that, by Anthropic's account, got Fable 5 to flag a few software flaws and in one case to write code showing how a flaw could be abused. Amazon CEO Andy Jassy escalated the findings directly to Treasury Secretary Scott Bessent, Commerce Secretary Howard Lutnick, and National Cyber Director Harry Coker Jr., and the export directive followed.

Commerce Secretary Howard Lutnick, who signed off on the reversal, said his department had 'worked closely with Anthropic to analyze and approve Fable 5.' In its own statement, Anthropic said Fable 5 returns globally on July 1, and by reporting from The Hacker News, Mythos 5 access had already been quietly restored on June 26 for roughly 100 US companies and federal agencies defending critical infrastructure. The negotiated conditions were substantive: proactive internal security research, early government access to future model launches before public release, reporting of any detected malicious usage, threat intelligence sharing, and participation in a cross-lab jailbreak severity framework with Amazon, Microsoft, and Google.

The honest caveat is what the reporting does not give you. It doesn't say exactly how the pre-release government access will work in practice, whether Anthropic's classifier that reportedly catches the jailbreak technique in over 99% of attempts holds up against adversarial red teams outside the company, or whether Commerce would issue an equivalent directive at OpenAI or Google if a similar partner-sourced finding surfaced there. What the resolution does establish is a template. Frontier labs now have a plausible playbook: patch the vulnerability, agree to pre-release review, help build a cross-industry severity framework, and get back to shipping. For federal buyers who want Mythos-class capability inside government workloads, that template is the actual product.