After two weeks of tense negotiations and a lockdown that left thousands of users without access to one of the world's most advanced artificial intelligence models, the Trump administration appears to have softened its stance toward Anthropic. Claude Mythos 5 is available again, although not for everyone or under the same conditions as before. Meanwhile, Claude Fable 5 is still waiting in the waiting room, with no clear signs of when it might return.

It should be remembered that on June 12, the United States Department of Commerce issued an export control directive that forced Anthropic to suspend access to its two most powerful AI models. The order cited national security reasons and required cutting off access to any foreign national, including the company's own non-U.S. employees. Given the impossibility of applying this restriction selectively without breaking the service, Anthropic chose to turn off both models for all its customers.

Why Trump stopped Claude's most powerful models cold

Washington's concern was not arbitrary. According to sources close to both the company and the administration, the directive was motivated in part by suspicions that a group linked to the Chinese government had managed to access the capabilities of Claude Mythos and its mass-market version, Fable 5. These models, released just days before, stand out for their ability to detect and exploit cybersecurity vulnerabilities with unprecedented speed and precision, which makes them tools with truly delicate dual-use potential.

In addition, Amazon CEO Andy Jassy would have directly expressed his fears to government officials about the potential of these AIs to generate information that can be used in cyber attacks. Pressure came from several fronts and the Trump administration was quick to act. Anthropic received the directive on a Friday at 5:21 PM and had no choice but to comply. The result was an unexpected blackout that affected hundreds of thousands of users just in the week of launch.

Claude Mythos 5 returns, but only for a select few

What changed this week is that the administration adjusted its stance. A letter dated June 26, signed by Secretary of Commerce Howard Lutnick and addressed to Anthropic co-founder Tom Brown, who led the negotiations, indicates that there has been a “review of licensing requirements” based on the work Anthropic did with the government to address the risks associated with Mythos 5 and Fable 5.

The practical result is that Mythos 5 can work again, but only for a small, approved group of organizations. Anthropic spokesperson Danielle Ghiglieri confirmed that the company has received notice that Mythos 5 may be redeployed to a small group of cyber defenders and infrastructure providers. In essence, the government did not lift the original directive, but rather created an exception for certain trusted organizations, the same scheme it previously applied with OpenAI's GPT-5.6 model, announced that same day. Anthropic now has basically the same deal as its rival.

The list of authorized organizations is not small in weight, although it is in number. The Project Glasswing program, which serves as a framework for controlled access to Mythos, includes names such as AWS, Apple, Cisco, CrowdStrike, Google, Microsoft and NVIDIA, among others, since its previous version. Under the new agreement, even non-U.S. employees at these organizations have the green light to use Mythos 5, which represents a significant relaxation from the harshness of the original directive.

Claude Fable 5 is still in limbo, and that matters

The problem is that Fable 5, the version of Mythos designed for the general public, still has no return date. While Mythos 5 was always intended for a closed circle of companies and agencies, Fable 5 was Anthropic's bet to bring those advanced capabilities to its paying subscribers and enterprise clients. Its prolonged absence has a real impact on hundreds of thousands of users who paid to access the best Anthropic had to offer and must now make do with older versions like Opus 4.8.

In his letter, Lutnick was explicit in noting that “all other requirements of the June 12 letter remain in effect until further notice” and that he reserves the right to reevaluate and adjust the scope of the licensing requirements if circumstances change. That is to say, nothing is guaranteed and the dispute between Washington and Anthropic is far from being completely resolved.

What this episode makes clear is that the intersection between advanced artificial intelligence, geopolitics and national security is no longer a theoretical debate. Claude Mythos 5 is not just a sophisticated chatbot; It is technology that governments consider powerful enough to regulate as if it were weapons. Anthropic, for its part, continues to bet on a negotiated exit that will allow it to expand access to Mythos 5 and finally return Fable 5 to the general market. How long that negotiation will take is, for now, the big unanswered question.