Anthropic's Mythos 5 Partially Relaunched After Negotiations

Mythos 5 from Anthropic Partially Relaunched After Negotiations

After a tumultuous two-week negotiation process with the Trump administration, Mythos 5 from Anthropic is finally back in action — at least partially, for a selected group of organizations, according to a letter from the government to Anthropic reviewed by The Verge. In contrast, Fable 5, the Mythos-class model intended for the general public, still appears to be on hold, with no apparent timeline for a deployment agreement.

The letter, dated June 26 and sent by Commerce Secretary Howard Lutnick to Anthropic co-founder Tom Brown, who recently led the negotiations, indicates that there has been a “revision of licensing requirements” due to Anthropic's recent “collaboration with the U.S. government to address the risks” associated with Mythos 5 and Fable 5.

Anthropic spokesperson Danielle Ghiglieri told The Verge in a statement that the company had “received notification from the U.S. government that Mythos 5, our most powerful cybersecurity model, can be redeployed to a small group of cybersecurity defenders and infrastructure providers.” She added, “We are working to provision all approved providers and restore their access to Mythos 5 as quickly as possible. We are excited 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.”

The U.S. government has not lifted the export control directive it imposed on Anthropic two weeks ago, prohibiting any foreign national from accessing either model (including Anthropic's own employees). Instead, the government has essentially made an exception for Mythos 5, approving a selected group of organizations for access, similar to what it did for GPT-5.6 from OpenAI, announced earlier in the day. According to the letter, Anthropic employees who are not U.S. citizens and members of the approved organizations who are not U.S. citizens are all allowed to access Mythos 5.

“These efforts have produced significant progress,” Lutnick wrote. “Furthermore, Anthropic has committed to working with the U.S. government on protocols, standards, and publications for [Mythos-class models]. In light of this progress... I have determined that appropriate safeguards are in place to allow certain trusted partners to access the Claude Mythos 5 model.”

Pressure has intensified on the Trump administration to modify the type of case-by-case regulatory environment it has recently adopted, especially as cybersecurity-focused models from its competitors continued to improve — and even outpace some cybersecurity-focused benchmarks. Pressure was also mounting within the AI industry in the U.S., particularly due to concerns about the advancements in AI that China might achieve while leading U.S. AI labs were sidelined. Additionally, there was the fact that leading U.S. government departments, such as the National Security Agency, had lost access to Mythos 5.

Now, Anthropic essentially has the same agreement as OpenAI: to release the model in a limited preview framework, only for approved organizations, such as trusted companies and the U.S. government itself. Both AI labs hope that more general availability will come soon, both for enterprise agreements and public access (as with Anthropic's Fable 5), but this will ultimately depend on the Trump administration. While many have called for AI regulation, including the AI labs themselves, a current consensus among some tech leaders is that the last two weeks have not been the right way to proceed.

In fact, OpenAI wrote in the company's blog post about GPT-5.6, “We do not believe that this type of government access process should become the long-term norm. It prevents the best tools from reaching the users, developers, businesses, cybersecurity defenders, and global partners who need them. We are taking this short-term measure because we believe it is the most solid path toward broader availability in the coming weeks, while working with the administration to develop the framework for the Executive Order on cybersecurity and a repeatable process for future model releases.”

Lutnick wrote in his letter to Anthropic, “All other requirements from the June 12 letter remain in effect until further notice.” He added, “I reserve the right to reassess and adjust the scope of licensing requirements on [Mythos 5 and Fable 5], if circumstances change.”