Trump Administration Partially Lifts Restrictions on Anthropic's Claude Mythos 5

The US government has eased export controls on Anthropic's most powerful AI model, Claude Mythos 5, allowing over 100 US organizations to regain access. The move follows weeks of tense negotiations between Anthropic and the Trump administration, though broader restrictions — including those on the consumer-facing Claude Fable 5 — remain in place.

The US government has partially lifted restrictions on Anthropic's Claude Mythos 5, permitting the company to restore access for more than 100 US organizations, including large corporations and government agencies.

The Government's Decision

In a letter obtained by WIRED, Commerce Secretary Howard Lutnick told Anthropic it could grant access to trusted partners because he had "determined that appropriate safeguards are in place."

"Anthropic has worked with the U.S. government to address risks associated with the Covered Models. These efforts have yielded significant progress." — Howard Lutnick, US Commerce Secretary

The letter also confirmed that organizations approved for Mythos access may now allow foreign national employees to use the model — a key reversal from the original June 12 directive, which had barred foreign nationals from accessing the models, even those living and working in the US.

What Remains Restricted

The easing is far from a full reinstatement. Key restrictions still apply:

Claude Fable 5 — the consumer-facing version of Mythos — remains unavailable for general use
All other requirements from the original June 12 directive remain in effect
No timeline has been announced for a broader rollout

"We are pleased to see this progress and continue to work with the government to expand access to Mythos 5 and make Fable 5 available for general use again." — Eduardo Maia Silva, Anthropic spokesperson

How the Crisis Unfolded

The Trump administration's intervention was triggered by a series of escalating concerns. Anthropic had granted access to a South Korean telecom firm the government believed had ties to China. Separately, Amazon and the National Security Agency raised alarms that Fable 5 could be jailbroken.

In response, Anthropic dispatched senior members of its cybersecurity and AI safety teams to Washington, DC. The company's negotiations with the US Department of Commerce were led by cofounder Tom Brown and public policy chief Sarah Heck.

Broader Implications for US AI Policy

The episode is more than a single company's regulatory headache — it's reshaping how frontier AI developers think about government relations.

Dean Ball, head of strategic futures at OpenAI and a former White House AI adviser, wrote in a June 16 blog post that the saga signals frontier model developers now "need an explicit green light from the government."

The ripple effects are already visible. OpenAI announced on Friday it was delaying the release of its upcoming GPT-5.6 models at the Trump administration's request.

A Costly Fight for Anthropic

The standoff has taken a measurable toll on the startup's business:

Anthropic sued the Trump administration earlier this year over a supply chain risk designation tied to its efforts to restrict how military contractors could use its AI
Following the June 12 directive, Anthropic investors worked through the weekend to assess the implications for the company's future
Access to Mythos 5 — described as the company's strongest cybersecurity model — was suspended during a critical period of commercial expansion

The partial reinstatement is a step forward, but it leaves unresolved questions about how aggressively the US government intends to regulate future AI model releases.