The National Security Agency (NSA) is reportedly utilizing Mythos Preview, a cutting-edge model developed by Anthropic that has not been made publicly available. This revelation comes from Axios and follows the Department of Defense's recent designation of Anthropic as a "supply chain risk" due to the company's refusal to grant unrestricted access to its capabilities.
Mythos, introduced earlier this month, is designed specifically for cybersecurity applications. Anthropic has expressed concerns regarding its potential for offensive cyber operations, leading to a decision to limit access to approximately 40 organizations, of which only a select few have been publicly acknowledged.
According to reports, the NSA is among the undisclosed entities with access to Mythos, primarily employing it to identify vulnerabilities within various environments. Furthermore, the UK's AI Security Institute has confirmed its access to this advanced model.
The expanding adoption of Anthropic's technologies by the U.S. military coincides with ongoing legal arguments asserting that these tools could pose risks to national security. The Pentagon's contention arose when Anthropic declined to make its AI model, Claude, available for widespread domestic surveillance and the development of autonomous weaponry.
Interestingly, the NSA's engagement with Mythos appears to signal a thawing relationship between Anthropic and the current administration. Recently, Anthropic's CEO, Dario Amodei, held a meeting with White House Chief of Staff Susie Wiles and Secretary of the Treasury Scott Bessent, which was described as productive.
As the NSA continues to explore the capabilities of Mythos, the implications for future cybersecurity measures and the evolving landscape of AI in defense are profound. This development may pave the way for enhanced security protocols and innovative applications in safeguarding critical infrastructure.