NSA Reported to Be Using Anthropic's Restricted Mythos AI Model Despite Ongoing Pentagon Tensions

The United States National Security Agency has reportedly begun using Anthropic's Mythos artificial intelligence model, a restricted system not publicly available, according to reporting published on April 20, 2026. The disclosure comes amid a period of significant friction between the Trump administration and Anthropic, which administration officials had previously labeled a politically radical organization and a potential national security concern.

The use of Mythos by the NSA is notable on several levels. First, it suggests that elements of the U.S. intelligence community have determined that Anthropic's AI capabilities offer operational value sufficient to override the political tensions that have characterized the company's relationship with the current administration. Second, it indicates that Anthropic has developed or made available a model tier beyond its publicly documented Claude product line, designed for use in sensitive or classified contexts.

Anthropic has not publicly acknowledged the existence of Mythos or confirmed the NSA's reported use of the system. The company has previously discussed its work with government and defense customers in general terms, noting that its safety-focused approach to AI development is intended in part to make its systems trustworthy for high-stakes applications. A restricted model offering would be consistent with the approach taken by other AI companies, which have developed government-specific versions of their systems to meet federal data handling and security requirements.

The reported adoption by the NSA stands in tension with public statements from other parts of the administration. The Trump White House had previously described Anthropic in pointed terms, calling into question the company's reliability as a technology partner for U.S. government agencies. The apparent disconnect between political rhetoric and operational practice within the federal government is not unprecedented in the history of technology procurement, where agencies frequently make acquisition decisions based on capability assessments that operate independently of executive branch communications.

The situation also highlights the complexity of Anthropic's position in Washington. The company has invested in policy engagement and has emphasized its commitment to AI safety as a feature relevant to national security. The reported use of Mythos by the NSA may represent early signs that those arguments are gaining traction within at least some parts of the intelligence and defense establishment, even as broader political tensions with the administration remain unresolved.

Sources: https://techcrunch.com/2026/04/20/nsa-spies-are-reportedly-using-anthropics-mythos-despite-pentagon-feud/