The two week ban on Anthropic's Mythos 5 and Fable 5 frontier models showed the Commerce Department can shut a frontier AI lab out of foreign markets with a single license call. The reversal helps Anthropic, and gives the government a working template for the next fight.
The U.S. government's two-week experiment with restricting Anthropic's most powerful AI models ended on Wednesday. The episode is worth studying less for what it cost Anthropic, which got its access back, and more for what it gave Washington: a working playbook for pressuring any frontier AI lab operating in foreign markets.
The Commerce Department lifted its export-control directive on Anthropic's Mythos 5 and Fable 5 models effective July 1, 2026, removing a license requirement that had blocked both models from being made available to anyone outside the United States. Commerce Secretary Howard Lutnick announced the reversal publicly on X, and Anthropic confirmed the change on its news page (Al Jazeera).
The original directive, issued by the Trump administration in mid-June, had cited national security concerns over an alleged jailbreak of Fable 5. Its scope was unusually broad: every foreign national, including Anthropic's own employees abroad, was cut off from the two models, forcing the company to abruptly disable Mythos 5 and Fable 5 for all customers worldwide (Wired; The Guardian).
Anthropic pushed back. In a public statement, the company said the alleged jailbreak targeted minor, previously known vulnerabilities that other public models can also surface, and that Fable 5's safeguards had passed thousands of hours of red-teaming with the U.S. government, the U.K. AI Safety Institute, and outside parties. No universal jailbreak had been found, the company argued (Anthropic).
The reversal came after Anthropic agreed to keep working with the government on security standards and to monitor for malicious use, terms the Commerce Department has not yet published in detail (BankInfoSecurity). The episode runs alongside Anthropic's separate lawsuit against the Department of Defense over its procurement of AI tools, a fight the company had opened before the export curbs landed. Co-founder Tom Brown is credited with improving the relationship with the Trump administration, though neither side has detailed what that involved (WJDX wire).
The interesting question is what happens the next time a frontier model goes viral for the wrong reason. Reuters and Axios, as carried by the WJDX wire, framed the original directive as a potentially dangerous precedent for AI regulation and as evidence of a shift in the Trump administration's posture from laissez-faire AI policy toward active oversight. Both warnings now have a working example attached. The Commerce Department has shown it can move fast, cite national security, and shut a lab out of its foreign user base on a single license call. The reversal only confirms the lever works: apply it, and a company with no domestic-only future has strong reasons to negotiate.
The remaining uncertainty is whether the next frontier release will need a quiet sign-off from Washington before launch, or whether the Anthropic case is a one-off negotiated outcome. Anthropic said in a follow-up post on Fable 5 that it was redeploying the model under new arrangements; it did not describe what those arrangements require in practice. For every other AI lab selling into Europe, Asia, and the Middle East, that is the line to watch.