NSA reportedly uses Anthropic’s Mythos despite Pentagon blacklist, exposing a split inside the U.S. government

The National Security Agency is reportedly using Anthropic’s Mythos Preview even though the Pentagon formally labeled Anthropic a supply-chain risk last month. Axios first reported the NSA use, citing two sources, and Reuters followed with a separate report saying it could not independently verify the claim.

If accurate, the report shows a clear gap between formal policy and operational practice. The NSA sits inside the Department of Defense, so any continued use of Mythos would mean at least part of the national security apparatus still sees value in Anthropic’s most sensitive cyber-focused model despite the official crackdown.

BEST SPRING 2026 DEALS

Editor's Choice

Private Internet Access

Access content across the globe at the highest speed rate.

70% of our readers choose Private Internet Access

70% of our readers choose ExpressVPN

ExpressVPN

Browse the web from multiple devices with industry-standard security protocols.

Nord VPN

Faster dedicated servers for specific actions (currently at summer discounts)

The timing matters because the dispute between Anthropic and the Pentagon remains active. The Pentagon informed Anthropic in early March that it had been designated a supply-chain risk, and Anthropic responded with a lawsuit arguing that the decision was unlawful and retaliatory.

How the Pentagon-Anthropic fight escalated

Anthropic’s relationship with the Defense Department did not start as a hostile one. In July 2025, the company announced a two-year Department of Defense agreement with a ceiling of $200 million to prototype frontier AI capabilities for national security use.

That relationship later broke down over how broadly Claude could be used in military settings. Reuters reported that Anthropic refused to remove two narrow restrictions, including use for mass domestic surveillance and fully autonomous weapons without human oversight, and the Pentagon later moved to blacklist the company from defense contracts.

Anthropic has argued that the restriction is narrower than many headlines suggest. In a March 5 statement, CEO Dario Amodei said the designation applies to use tied directly to Department of Defense contracts, not to every Anthropic customer or every government-related use case.

Why Mythos is different

Mythos Preview is not an ordinary enterprise chatbot. Anthropic introduced it this month as a tightly restricted model with unusually strong cyber capabilities, especially around identifying and testing software vulnerabilities.

Reuters said concerns around Mythos center on its ability to accelerate offensive cyber work as well as defensive security testing. Anthropic itself has limited access to a small group of organizations, and public reporting says the model is not broadly available because of those risks.

That makes the NSA report more significant than a simple procurement contradiction. If the agency is using Mythos while the Pentagon still treats Anthropic as a supply-chain risk, it suggests national security demand for advanced cyber AI may be stronger than the government’s own public restrictions.

What we know, and what remains unconfirmed

The strongest confirmed point is that Anthropic and the Pentagon are in an active legal and political fight. Reuters, Anthropic’s own statements, and court reporting all show that the blacklisting happened and that Anthropic sued to challenge it.

The NSA use claim is less settled. Axios reported it first, Reuters repeated the report, and Reuters said the NSA, the Department of Defense, and Anthropic did not immediately respond to requests for comment at the time. That means the claim carries real weight, but it still rests on sourced reporting rather than an on-the-record government confirmation.

There are also signs that relations may be thawing. Reuters reported that Anthropic CEO Dario Amodei met with White House officials on April 17, and President Trump said on April 21 that Anthropic was “shaping up” and that a Pentagon deal could still happen.

Key points

Why this story matters

This is really a governance story as much as an AI story. A government cannot credibly call a company a national security supply-chain risk while parts of that same system reportedly keep using its most sensitive model, unless there is an exemption, a workaround, or a policy split that officials have not explained publicly.

It also shows how fast cyber-capable AI has moved from product launch to state use. Mythos only surfaced publicly this month, yet it is already at the center of procurement fights, court battles, access controls, and reported intelligence adoption.

For now, the cleanest way to describe the situation is this: the blacklist is real, the legal fight is real, and the reported NSA use remains credible but not officially confirmed. That distinction matters, especially in a story involving current U.S. national security policy.

FAQ