Axios reveals US security agency using blacklisted AI model Mythos

A new report from Axios has raised serious questions about how advanced AI tools are being used inside the U.S. government. According to the report, the National Security Agency has been using an AI model called Mythos Preview, even after concerns were flagged by the Pentagon.

The model comes from Anthropic, a major player in the artificial intelligence space. What makes this situation controversial is that the company had already been given a formal supply chain risk designation by the Department of Defense, which typically signals caution or restricted use.

What the “blacklist” designation actually means

A supply-chain risk designation does not always amount to a government-wide ban, but it can impose serious restrictions.

In Anthropic’s case, Reuters reported that the Pentagon’s action followed a dispute over guardrails on military use and barred Pentagon contractors from using the company’s technology in military work.

Despite that, the Axios report claims the Mythos Preview model was being used more broadly within defense circles. Reuters noted it could not independently verify the report, and officials from Anthropic, the NSA, and the Department of Defense did not immediately comment.

US National Security Agency department — *Source: visuals6x/Depositphotos*

Little-known fact: The National Security Agency was established in 1952 and is primarily responsible for signals intelligence and cybersecurity operations for the United States.

Why Mythos is drawing attention

The Mythos model is not just another AI chatbot. Anthropic has described it as its most capable system so far, especially when it comes to coding and agent-like behavior. That means it can perform complex tasks with limited human input, making it powerful but also potentially risky.

Experts have pointed out that tools like Mythos could identify software vulnerabilities at a very high level. In the wrong context, that same capability could be used to exploit systems instead of protecting them, which is why cybersecurity concerns are growing.

Rising fears around AI and cyberattacks

One of the biggest concerns tied to Mythos is its potential to accelerate cyberattacks. Advanced AI systems can analyze code faster than humans and uncover weaknesses that might otherwise go unnoticed.

Security experts worry that if such tools are not tightly controlled, they could be misused either internally or by adversaries. This has made governments more cautious about how and where these models are deployed, especially in national security environments.

A complicated relationship between Anthropic and the Pentagon

The broader relationship between Anthropic and the U.S. government has become increasingly complicated. Earlier tensions with the Pentagon centered on Anthropic’s refusal to relax guardrails that limited military uses of its AI, including for autonomous weapons and domestic surveillance.

Recently, however, there have been signs of renewed cooperation. Discussions between the Trump administration and Anthropic’s leadership suggest both sides may be exploring ways to work together again, even as concerns about Mythos continue.

What this means for AI oversight in government

This report highlights a growing challenge for governments worldwide. AI systems are advancing quickly, often faster than the policies designed to regulate them. That creates gaps where powerful tools may be used before clear rules are in place.

If the claims about Mythos are accurate, it could prompt stricter oversight and clearer guidelines on AI use within defense agencies. It may also lead to more scrutiny of how companies like Anthropic develop and distribute high-capability models.

Little-known fact: Anthropic was founded in 2021 by former OpenAI researchers and has focused heavily on building AI systems with safety and alignment as core priorities.

The bigger picture for AI in national security

The use of AI in national security is becoming unavoidable. Governments see these tools as essential for staying competitive, but they also come with significant risks. Balancing innovation with safety is turning into one of the biggest policy challenges of this decade.

The NSA flag painted on a brick — *Source: BreizhAtao/Depositphotos*

The Axios report adds another layer to that debate. Whether confirmed or not, it underscores how high the stakes have become when cutting-edge AI meets national defense.

TL;DR

Axios reports the National Security Agency used Anthropic’s Mythos AI despite Pentagon risk concerns.
Mythos received a Pentagon supply-chain risk designation that restricted use of its technology in Pentagon contracts and raised broader security questions.
Anthropic says Mythos is its most advanced model for coding and autonomous tasks.
Experts warn that it could identify and exploit cybersecurity vulnerabilities.
The situation highlights growing tension between AI innovation and national security oversight.

This article was made with AI assistance and human editing.

Don’t forget to follow us for more exclusive content.

If you liked this, you might also like: