Anthropic cleared to redeploy Mythos 5 to over 100 US

Anthropic has permission from the US Commerce Department to restore access to its most advanced model, Claude Mythos 5, for a limited group of more than 100 US organizations, including large corporations and government agencies. Commerce Secretary Howard Lutnick told Anthropic cofounder Tom Brown in a letter that he had “determined that appropriate safeguards are in place.”

What exactly did the Commerce Department authorize?

The department permitted Anthropic to grant Mythos 5 access to a narrow set of vetted partners, specifically more than 100 US organizations, and to allow foreign national employees of those organizations and Anthropic to use the model. The letter from Secretary Lutnick, obtained by WIRED, says the move reflects his determination that safeguards are adequate, but also reiterates that other requirements from the initial export directive sent on June 12 remain in effect.

Anthropic described Mythos 5 as “our strongest cybersecurity model” in a statement from spokesperson Eduardo Maia Silva, and said it is working to provision the approved providers and restore their access "as quickly as possible." The company also said discussions with the White House are ongoing about restoring access to Claude Fable 5, the consumer-facing variant released with additional safeguards.

Why did the government step in and restrict access in the first place?

US officials intervened after learning Anthropic had granted access to a South Korean telecommunications firm the White House believed had ties to China, and after separate concerns from Amazon and the National Security Agency that Fable 5 could be jailbroken. The White House sent an export control directive on June 12 that required limits on foreign-national access to Mythos and Fable 5, and Anthropic responded by disabling access to the models entirely.

Anthropic subsequently sent senior members of its cybersecurity and AI safety teams, including Tom Brown and public policy chief Sarah Heck, to meet with Commerce and other Trump administration officials as it sought to resolve the restrictions. Semafor first reported the existence of Lutnick’s letter to Anthropic.

How are other companies reacting and what spillover has this caused?

The episode has affected other frontier AI developers: OpenAI announced it was delaying the release of its upcoming GPT 5.6 models after a request from the Trump administration. Former White House AI adviser Dean Ball, now head of strategic futures at OpenAI, wrote on June 16 that the dispute has shown developers they “need an explicit green light from the government now.” Anthropic also sued the administration earlier this year over a supply chain risk designation tied to how its models could be used by military contractors.

Why it matters

The Commerce Department’s partial reinstatement signals that the administration intends to exert granular control over who can run frontier models and under what safeguards. Developers now face a de facto requirement to secure government approval for broad rollouts, and companies that sell advanced models will need to make engineering and policy trade-offs to satisfy national security and export-control conditions.

This is not just a regulatory detail. The decision affects which organizations can use top-tier cybersecurity and infrastructure tools, who may have access to capabilities for defensive and offensive use, and how quickly consumer-facing variants like Fable 5 can return to general availability.

What to watch

Watch whether the White House and Anthropic reach a broader agreement on Fable 5 access and whether the Commerce Department publishes a clear, repeatable framework for future model releases. Also monitor whether other model makers receive similar, explicit approvals or are asked to delay launches, following OpenAI’s decision on GPT 5.6.