Anthropic’s AI model, known as Mythos, managed to breach most of the top-secret systems of the U.S. National Security Agency (NSA) in just a few hours. This information was published by The Economist, citing sources familiar with the matter.
“On June 11, Mark Warner, vice chair of the Senate Intelligence Committee, stated that General Joshua Rudd, who heads both the NSA and the Pentagon’s Cyber Command, had informed him that Mythos ‘compromised nearly all our classified systems not in weeks, but in hours,'” writes The Economist.
No further details have been disclosed regarding the specific nature or extent of the access gained by the system.
This revelation comes amid a strained relationship between Anthropic and the Pentagon. In March 2026, the U.S. Department of Defense labeled the company a “supply chain threat,” effectively ending their collaboration.
According to reports, the conflict stemmed from disagreements over the military application of artificial intelligence. The Pentagon sought the right to utilize Anthropic’s developments “for all lawful purposes,” but the company demanded that restrictions be imposed. For instance, Anthropic refused to permit the use of its models for mass surveillance of U.S. citizens or for the creation of fully autonomous lethal weapons systems.
About the Author
