US Government Partially Restores Access to Anthropic's Advanced Claude Mythos Model

The Trump administration has partially lifted export controls on Anthropic's most powerful AI model. Commerce Secretary Howard Lutnick sent a letter to Anthropic authorizing the company to restore access to Claude Mythos 5 for more than 100 US organizations, including corporations and government agencies deemed "trusted partners."

The move represents a modest thaw in what has been a contentious relationship between Anthropic and the White House over AI safety and national security. Earlier this month, the Trump administration issued an export control directive requiring Anthropic to restrict foreign nationals—including those working in the US—from accessing both Mythos 5 and its consumer-facing sibling, Claude Fable 5.

What Changed

Lutnick acknowledged that "appropriate safeguards are in place" following Anthropic's cooperation with government officials. The letter also now permits approved organizations to grant access to their foreign national employees, a significant loosening of the initial blanket restriction.

However, the partial reinstatement stops well short of unrestricted deployment. The White House remains silent on when—or if—Claude Fable 5 will return to general public availability.

The Broader Context

The saga reflects widening tensions over AI governance. Government concerns arose after Anthropic provided access to a South Korean telecommunications firm with suspected ties to China. Amazon and the NSA subsequently raised alarms that Fable 5 could be jailbroken, spurring the administration to act.

Anthropic dispatched senior officials from its cybersecurity and safety teams to negotiate with the Commerce Department. The gambit appears to have yielded limited gains: Mythos 5 for vetted defenders and infrastructure providers, but broader public access remains restricted.

Industry-Wide Implications

The standoff carries consequences beyond Anthropic. OpenAI recently delayed its GPT 5.6 release following a similar Trump administration request, signaling a broader shift toward pre-release government approval for frontier models. As one former White House AI adviser noted, developers now understand they "need an explicit green light from the government" before deploying powerful systems.

For Anthropic, the battles have proven costly—the company sued the administration over supply chain restrictions and faced investor uncertainty after the June 12 directive.