Trump admin grants Anthropic export license for Mythos 5, ending 2-week standoff

The Trump administration granted Anthropic an export license on Friday, June 26, 2026, allowing the company to release its Mythos 5 model to a group of roughly 100 companies and federal agencies after a two-week export control standoff. Commerce Secretary Howard Lutnick issued a letter confirming the decision, marking a major resolution to the conflict that began on June 12 when the administration abruptly issued an export control directive suspending all access to Fable 5 and Mythos 5 by foreign nationals, citing national security concerns over a claimed jailbreak.

The dispute stemmed from Amazon and other companies flagging a potential vulnerability in Fable 5—the ability to read a codebase and identify security flaws—that they characterized as a national security risk for offensive cyber operations. Anthropic disputed the severity, arguing the technique was already available in other models like GPT-5.5 and that the finding of a narrow, non-universal jailbreak did not justify pulling access to models deployed to hundreds of millions of people. Senior Anthropic staff met with Trump administration officials multiple times, including at the G7 summit in Évian-les-Bains, where Trump reportedly eased concerns about the company's models.

The license authorizes release to a curated set of US companies and agencies, not the full rollout, but it unblocks Mythos 5 for enterprise partners after Anthropic lost access across its user base. This resolution reflects a softer stance from the administration than the initial embargo, though the precedent of export controls on specific frontier models remains a defining uncertainty for US AI companies.

For architects: regulatory control over model access is now operational. Whether export restrictions become routine or remain exceptional will shape deployment and partnership strategies around frontier AI inference and reasoning workloads.