The United States government has granted Anthropic permission to release its highly capable Mythos artificial intelligence model to a restricted group of organizations and companies. This decision follows a revision of licensing requirements, which occurred after federal authorities ordered an export block earlier this month due to national security concerns.
Commerce Secretary Howard Lutnick addressed the company in a letter dated Friday, noting that since the export prohibition was issued in June, Anthropic has worked closely with the federal government to mitigate risks regarding these covered models. Lutnick stated that, in view of the progress made during these efforts, he has determined that sufficient safeguards are now in place to allow designated trusted partners to access the Claude Mythos 5 model.
However, the authorization does not extend to the release of Fable, a version of the technology that is less powerful than Mythos. Anthropic confirmed in an official statement that they received notice from the government allowing the redeployment of Mythos 5—described by the company as their most advanced cybersecurity-focused model—to a limited selection of infrastructure providers and cyber defenders.
The company noted that it is currently working to provision these approved partners and restore their access to Mythos 5 as swiftly as possible. They expressed satisfaction with the progress and indicated ongoing collaboration with the government to eventually broaden access to the Mythos 5 model and restore the general availability of Fable 5. According to a source familiar with the discussions, negotiations between the government and Anthropic are expected to persist through the weekend, with a focus on restoring access to Fable as well.
Earlier this June, Anthropic had suspended all customer access to both Mythos and Fable to ensure compliance with a government mandate that halted use by foreign nationals, a requirement that also applied to Anthropic’s internal staff. These negotiations highlighted the absence of a unified regulatory approach toward artificial intelligence, even as the field evolves rapidly and the United States seeks to maintain its edge over international competitors like China.
While the administration led by President Trump has largely maintained a non-interventionist stance on AI regulation to help domestic firms lead globally, the recent actions against Anthropic—alongside requests for OpenAI to restrict its latest model—reflect growing concerns. Experts have argued for months that models like Mythos could pose security risks by enabling malicious actors to identify and exploit technical vulnerabilities with extreme speed.
This is not the first instance of friction between federal authorities and Anthropic. Earlier this year, the administration designated the firm a “supply chain risk” and effectively blacklisted it due to disagreements concerning military applications of the company’s technology. Anthropic has since filed a lawsuit challenging this designation, securing at least one preliminary victory in the legal proceedings.
