Anthropic Launches Secret Claude Mythos Model for US National Security Agency

A major controversy is brewing around Anthropic, the American company that produces Claude artificial intelligence models. Citing sources familiar with the situation, the Financial Times reports that the US National Security Agency (NSA) is using a specialized Claude Mythos model for cyber operations tasks. This occurs despite the fact that cooperation between the company and US Department of Defense entities was officially banned. Ixbt.com reports .

According to the publication, Anthropic engineers have been working with NSA divisions in recent months, helping to adapt the Mythos model for specialized tasks. While officially the focus is on cybersecurity needs, sources suggest that this technology could be used in offensive operations, such as identifying and exploiting vulnerabilities in foreign computer networks.

Modern cyber conflicts are evolving much faster than they did a few years ago. If it previously took months between identifying a vulnerability and exploiting it, today this time is measured in hours or even minutes. In such conditions, artificial intelligence tools are becoming a crucial advantage for intelligence agencies. The Claude Mythos neural network is capable of automatically analyzing software and finding numerous vulnerabilities in operating systems and browsers.

The conflict between the company and the US administration began in early 2026, when the White House was negotiating a $200 million contract with Anthropic. At that time, the government demanded the relaxation of restrictions on AI systems, but company CEO Dario Amodei opposed the use of AI in mass surveillance and autonomous combat systems. As a result, in March 2026, the Pentagon declared Anthropic products a "systemic risk" to national security.

Currently, the company is defending its rights in court, while government agencies cite national security concerns. If the Financial Times information is confirmed, the situation appears paradoxical: special services continue to secretly use technologies from a company that is in dispute with the federal government and has been officially deemed dangerous.