NSA uses anthropic myth despite Pentagon dispute Apfelpatient

According to a report by Axios, the US National Security Agency (NSA) is using Anthropic's new cybersecurity model Mythos Preview – even though the Trump administration had actually kept Anthropic away from US authorities.

The relationship between Anthropic and the US government has been strained for months. In February, President Trump ordered all government agencies to cease using Anthropic services after the company refused to relax certain security measures for military use during contract negotiations. Simultaneously, Anthropic sued the Department of Defense after the Trump administration classified the company as a "supply chain risk." Now it has emerged that the NSA is nevertheless using Anthropic's brand-new cybersecurity model, Mythos.

The Axios report is based on two sources familiar with the matter. According to these sources, the NSA is among the approximately 40 organizations that Anthropic has granted access to Mythos Preview. One of the sources even stated that the model is being used more extensively within the agency. This is noteworthy because Mythos was only introduced in early April as part of Project Glasswing – a cybersecurity initiative in which Apple is also involved.

What Claude Mythos can do

Anthropic describes Mythos as a powerful general-purpose model that performs particularly well in computer security tasks. The model is not publicly available, but only accessible to select partners – including Apple, Google, Microsoft, Nvidia, and apparently now also the NSA.

Unlike OpenAI's comparable cybersecurity model, GPT-5.4 Cyber, Anthropic pursues a more collaborative approach with a broad coalition of technology companies and government agencies. The fact that the NSA is among its users underscores the importance that Anthropic's model apparently holds in the field of cyber defense.

The meeting at the White House

The information about the NSA use comes shortly after a much-discussed visit: Anthropic CEO Dario Amodei met on Friday with White House Chief of Staff Susie Wiles and other government officials. According to a Reuters report, the meeting was supposed to focus on Mythos. The White House described the meeting as "productive and constructive," while President Trump, when questioned by reporters, stated that he had "no idea" about it.

The discrepancy shows that the relationship between Anthropic and the Trump administration remains complex – despite open legal conflicts, practical cooperation apparently takes place at the working level.

Anthropic: The ongoing legal dispute

The legal battle continues. In March, Anthropic filed lawsuits against the Department of Defense in two courts after the Trump administration classified the company as a "supply chain risk." One court granted Anthropic a preliminary injunction temporarily blocking this classification. Another court, however, denied the request.

For Apple and other technology partners who work closely with Anthropic, this development is interesting: Even though the political climate in the US is tense, the technological collaborations continue. Apple itself continues to work with Anthropic on Project Glasswing to test iOS and macOS for vulnerabilities and strengthen the cybersecurity of its own products.

Why this goes beyond the NSA

The fact that the NSA is using Mythos demonstrates two things: First, how much demand there is now for specialized AI models for cybersecurity, even in the government sector. Second, that even in situations of political tension between companies and governments, technological dependence often overshadows political considerations. This is rather good news for the further development of AI-supported cybersecurity – even if the legal questions surrounding Anthropic and the Pentagon remain unresolved. (Image: Shutterstock / gguy)