After an intense two-week negotiation process with the Trump administration, Anthropic’s advanced cybersecurity model, Mythos 5, is now approved for limited use by a select number of organizations. A letter from Commerce Secretary Howard Lutnick to Anthropic co-founder Tom Brown confirmed the revised license requirements following Anthropic’s collaboration with the U.S. government to mitigate risks associated with both Mythos 5 and its public-facing counterpart, Fable 5.
Danielle Ghiglieri, a spokesperson for Anthropic, expressed satisfaction with the government’s decision, stating that Mythos 5 can now be redeployed to a small group of cyber defenders and infrastructure providers. The company is currently working to restore access for these approved organizations promptly. However, the public release of Fable 5 remains uncertain, with no timeline provided for its general availability.
Despite the positive development regarding Mythos 5, the U.S. government maintained its export control directive initially imposed on Anthropic. This restriction prevents any foreign nationals, including Anthropic employees, from accessing either model. An exception has been granted for Mythos 5, allowing certain organizations, similar to the recent access permitted for OpenAI’s GPT-5.6, to utilize the technology.
In the letter, Lutnick noted significant progress in negotiations and mentioned Anthropic’s commitment to continue working alongside the U.S. government on protocols and standards for Mythos-class models. He emphasized that adequate safeguards are now in place, permitting trusted partners to access Mythos 5.
The regulatory environment has come under scrutiny, especially as competitors in the AI space enhance their cybersecurity models, raising concerns over the stagnation of U.S. advancements compared to potentially growing capabilities in other countries, notably China. High-profile government departments, including the National Security Agency, had previously lost access to Mythos 5, adding urgency to negotiations.
Anthropic’s current arrangement mirrors that of OpenAI: a limited preview release for approved organizations, which includes trusted enterprises and U.S. government entities. Both companies are hopeful for broader general availability soon, although the final decision rests with the Trump administration.
Despite calls for more structured AI regulations, some leaders within the tech industry have expressed that the current method may not be ideal. In a recent blog post regarding GPT-5.6, OpenAI stated their belief that this approach should not become the long-term standard, as it restricts access to critical tools for users, developers, and various organizations that require these capabilities for cybersecurity and other applications.
Lutnick’s letter reiterated that all previous conditions remain effective until further notice, with a caveat that he could adjust licensing scope for Mythos 5 and Fable 5 if circumstances change.
