NSA spies are reportedly using Anthropic’s Mythos, despite Pentagon feud

Reports indicate that the National Security Agency (NSA) is utilizing Mythos Preview, Anthropic's newly unveiled model that has not been made publicly available. This revelation comes shortly after the Department of Defense, which oversees the NSA, categorized Anthropic as a potential 'supply chain risk' due to the company's refusal to grant Pentagon officials unrestricted access to the model's capabilities. Mythos, introduced by Anthropic earlier this month, is designed specifically for cybersecurity tasks, yet the firm has expressed concerns over its potential for offensive cyber operations, leading to a decision to restrict its public release. With access limited to around 40 organizations, only a small number of which have been disclosed, the NSA is believed to be among those using Mythos to identify vulnerabilities in various environments. Additionally, the UK’s AI Security Institute has confirmed its own access to Mythos, highlighting its growing importance in cybersecurity efforts. The U.S. military's increasing reliance on Anthropic's AI tools coincides with ongoing legal disputes where the Pentagon argues that these technologies pose risks to national security. Tensions escalated when Anthropic declined to make its Claude model available for extensive domestic surveillance or for use in developing autonomous weapon systems. Meanwhile, it seems that Anthropic's relationship with the current administration is improving. Recently, CEO Dario Amodei met with key White House officials, including Chief of Staff Susie Wiles and Secretary of the Treasury Scott Bessent, with reports indicating that the meeting was constructive. TechCrunch has sought comments from the NSA, but Anthropic has opted not to provide any statements.