Cybersecurity Policy Report, Administration Allows Limited Release of Anthropic’s Cyber-Capable AI Model, (Jun 29, 2026)
The Trump administration has eased restrictions that had prevented Anthropic from publicly releasing Mythos 5, an AI (artificial intelligence) model that the administration was concerned could be used by foreign adversaries for cyber attacks and other malicious activities.
In a letter sent Friday, Commerce Secretary Howard Lutnick told Tom Brown, Anthropic’s chief compute officer, that the company could give “certain trusted partners” access to Mythos 5.
The Commerce Department had imposed export controls on the Mythos 5 and Fable 5 models to prohibit their use by foreigners. Anthropic responded by canceling the public release of the models because it was unable to determine which users might be subject to the export restrictions (CPR, June 15).
Fable 5 and Mythos 5 are based on Anthropic’s Claude Mythos2 Preview model, which proved to be so adept at finding and exploiting cyber vulnerabilities that Anthropic withheld it from public release. Those hacking capabilities were stripped from the publicly released Fable 5 and Mythos 5 models, but the administration had lingering concerns that the models could be misused by foreign adversaries.
Those concerns have abated, at least for Mythos 5, Mr. Lutnick said. “Anthropic has worked with the U.S. government to address risks associated with the Covered Models. These efforts have yielded significant progress. In addition, Anthropic has committed to work with the U.S. government on protocols and standards and releases for the Covered Models,” he told Mr. Brown.
“In light of this progress, as well as the Department of Commerce’s (‘Commerce’) evaluation of the diversion risks currently presented by the Covered Models, I have determined that appropriate safeguards are in place to permit certain trusted partners to access the Claude Mythos 5 Model,” he said.
Mr. Lutnick’s letter included an annex, which was not publicly released, that named entities to which Anthropic could provide access to Mythos 5, including Anthropic’s own “foreign national employees,” without obtaining an export control license.
“Commerce may amend the list of entities in Annex A at any time,” Mr. Lutnick said. “I reserve the right to reevaluate and adjust the scope of license requirements on the Covered Models, should circumstances change.”
The Trump administration initially adopted a hands-off policy toward the development and release of “frontier” AI models but reversed course in an executive order that encouraged AI companies to voluntarily submit advanced models to an “AI cybersecurity clearinghouse” so that they could undergo national security evaluations (CPR, June 2).
The EO was followed, however, by the unexpected export controls imposed on Anthropic and a White House request to OpenAI that it delay the public release of its newest models until they underwent a national security assessment (CPR, June 26).
MainStory: TopStory FederalLegislation DataSecurity AINews