NSA Uses Anthropic Mythos AI Despite Pentagon Blacklist

Key Takeaways

• The National Security Agency (NSA) is reportedly using Anthropic’s advanced Mythos Preview AI model.
• This deployment occurs despite the Department of Defense (DoD) having designated Anthropic as a “supply chain risk” and a presidential order banning its technology.
• The Pentagon’s blacklist stems from Anthropic’s refusal to remove safety guardrails against mass surveillance and autonomous weapons from its DoD contracts.
• Mythos Preview is a highly restricted AI model, raising concerns due to its significant offensive cybersecurity capabilities.

The National Security Agency (NSA) is reportedly utilizing Anthropic’s sophisticated AI model, Mythos Preview, creating a stark contradiction within U.S. federal policy. This deployment directly challenges the Department of Defense’s (DoD) official classification of Anthropic as a “supply chain risk” and a presidential directive to cease using its technology. The ongoing situation highlights a significant internal conflict regarding AI governance within national security agencies.

The friction between Anthropic and the Pentagon emerged in early 2026. Initially, in July 2025, Anthropic secured a 200 million dollar contract with the DoD. This agreement included explicit stipulations prohibiting the application of its Claude AI for extensive domestic surveillance or the development of fully autonomous weapon systems.

The Pentagon-Anthropic Rift

The contractual relationship remained stable until January 2026. At that point, Defense Secretary Hegseth issued a memorandum mandating the inclusion of “any lawful use” clauses across all DoD AI contracts. This effectively demanded that Anthropic dismantle the previously agreed-upon safety restrictions.

Anthropic declined to comply. In response, by late February 2026, the Pentagon officially labeled the company a “Supply-Chain Risk to National Security.” This designation instructed all contractors and suppliers engaged with the U.S. military to immediately terminate commercial dealings with Anthropic.

Adding to the pressure, President Trump issued a separate directive ordering all federal agencies to discontinue the use of Anthropic’s technology, allowing for a six-month phase-out period for already integrated systems.

Despite these comprehensive bans, a recent Axios report published on April 19, 2026, exposed a striking inconsistency. The NSA, an agency operating under the purview of the Department of Defense, has been actively employing Mythos Preview, Anthropic’s latest and most potent AI model.

Sources cited by Axios indicate that Mythos Preview is being used “more widely” within the NSA, with its application reportedly extending to other segments of the department. Anthropic has severely restricted access to Mythos, granting availability to only approximately 40 organizations globally due to apprehensions regarding its potential for offensive cybersecurity operations.

Of these select organizations, only 12 have been publicly disclosed, and the NSA is reportedly among the undisclosed entities granted access. A primary concern fueling both the secrecy and controversy surrounding Mythos is its advanced capacity for offensive cyber operations. Anthropic itself has acknowledged fears that the model could significantly enhance cyberattacks if misused, which is a key factor in its tightly controlled distribution.

This makes the NSA’s adoption particularly noteworthy, given its role as the nation’s foremost signals intelligence and cybersecurity agency, operating under U.S. Cyber Command. In the United Kingdom, the NSA’s counterparts are reportedly accessing Mythos through the AI Security Institute, suggesting that allied intelligence communities are also discreetly integrating the model.

Legal Battle Continues

In March 2026, Anthropic initiated a lawsuit in San Francisco, characterizing the Pentagon’s supply-chain designation as “unprecedented and unlawful.” The company alleges violations of its free speech and due process rights. The legal proceedings are ongoing, even as the military simultaneously argues in court that using Anthropic tools poses a national security threat while internally deploying those very tools.

On April 17, Anthropic CEO Dario Amodei met with White House Chief of Staff Susie Wiles and Treasury Secretary Scott Bessent. Discussions centered on the government’s current use of Mythos, its associated security protocols, and a potential strategy for wider adoption across other federal departments.

The NSA’s quiet utilization of Mythos, while the Pentagon is locked in a legal battle with Anthropic, suggests that operational necessity is taking precedence over official policy within the U.S. intelligence community. As AI capabilities rapidly advance, the divergence between institutional prohibitions and on-the-ground implementation is widening, raising critical questions about governance, oversight, and accountability in the deployment of advanced AI within national security agencies.

Disclaimer: HackersRadar reports on cybersecurity threats and incidents for informational and awareness purposes only. We do not engage in hacking activities, data exfiltration, or the hosting or distribution of stolen or leaked information. All content is based on publicly available sources.