The US government has partially eased restrictions on Anthropic’s most advanced AI model, Claude Mythos 5, allowing access to return for a limited set of US organizations. The move gives Anthropic a path to redeploy Mythos 5 to more than 100 US organizations, including large corporations and government agencies, while leaving wider access controls in place.
The decision follows weeks of tension between Anthropic and the Trump administration over how powerful AI systems should be released, who should be allowed to use them, and what safeguards the government expects before access is restored.
What the government approved
In a letter to Anthropic’s cofounder and chief compute officer Tom Brown, Commerce Secretary Howard Lutnick said certain trusted partners could regain access to Mythos because he had “determined that appropriate safeguards are in place.” WIRED obtained the letter, and Semafor first reported on its existence.
Lutnick also wrote that Anthropic had worked with the U.S. government on risks tied to the Covered Models. In his words, “Anthropic has worked with the U.S. government to address risks associated with the Covered Models. These efforts have yielded significant progress.”
The approval is not a full reopening. The White House did not permit a broader roll out of Mythos, and it did not clarify what will happen to Claude Fable 5, the consumer-facing version of Mythos that Anthropic released with significantly more safeguards.
Lutnick also noted that all other requirements in his initial directive sent on June 12 remain in effect. That means the government’s broader intervention has not ended, even though a defined group of users can return to Mythos 5.
Who may regain access
Anthropic spokesperson Eduardo Maia Silva described the restored access as limited to a specific set of users. In a statement to WIRED, he said Mythos 5, “our strongest cybersecurity model,” can be redeployed to “a small group of cyber defenders and infrastructure providers.”
According to Silva, Anthropic is working to provision the approved set of providers and restore their access as quickly as possible. He also said the company continues to work with the government to expand access to Mythos 5 and make Fable 5 available for general use again.
The latest letter also changes the treatment of foreign national employees for approved organizations. Under the earlier directive, Anthropic was required to limit foreign nationals from accessing Claude Mythos and Fable 5, including people working and living in the United States.
Now, organizations approved to use Mythos may extend access to their foreign national employees. Anthropic’s foreign national employees may also receive access under the latest decision.
Why Mythos 5 became restricted
The Trump administration’s concerns grew after it learned Anthropic had shared Mythos access with a South Korean telecommunications firm that officials believed had ties to China, according to WIRED’s earlier reporting.
The issue expanded after Amazon and the NSA raised concerns that Claude Fable 5, a safeguarded version of Mythos 5, could be jailbroken. Those concerns helped convince the Trump administration that it needed to act.
In recent weeks, Anthropic sent senior members from its cybersecurity and AI safety teams to Washington, DC, to meet with Trump administration officials. Tom Brown has been part of that process, and public policy chief Sarah Heck has also been leading discussions with the US Department of Commerce.
The result is a narrow reopening rather than a clean reset. Mythos 5 can return to selected organizations, but the government is still keeping significant limits around Anthropic’s model releases.
What this means for AI releases
The partial reinstatement is an important business and policy development for Anthropic. Getting Mythos 5 back online marks progress for the company and the White House, but the dispute has also raised a larger question: how much control will the Trump administration seek over future frontier AI model releases?
The Anthropic case is not happening in isolation. On Friday, OpenAI announced it was delaying the release of its upcoming GPT 5.6 models in response to a request from the Trump administration.
Dean Ball, head of the strategic futures team at OpenAI and a former White House AI adviser, wrote in a June 16 blog post that Anthropic’s latest conflict with the Trump administration has shown frontier AI model developers they “need an explicit green light from the government now.”
For AI companies, the message is practical: model launches may increasingly depend not only on internal safety reviews and customer readiness, but also on direct government approval. The Mythos 5 decision shows that access can be restored when officials accept the safeguards, but it also shows that approval may be limited, conditional, and tied to specific users.
The unresolved questions
Several major questions remain open. The White House has not allowed a broader Mythos roll out. It has not said what will happen to Claude Fable 5. And the June 12 directive still remains in effect outside the access now restored for approved partners.
The fight has already been costly for Anthropic’s business. The company sued the Trump administration earlier this year over a supply chain risk designation it received after trying to set redlines around how military contractors could use its AI models.
WIRED also previously reported that some Anthropic investors worked over last weekend to determine what the White House’s restrictions on Claude Fable 5 meant for Anthropic’s corporate future.
For now, Anthropic has won back access for a defined group of organizations and employees. But the larger conflict over frontier AI control, safeguarded model releases, and government signoff remains unsettled.