The National Security Agency (NSA) is reportedly using Anthropic’s Mythos Preview, a restricted cyber model that Anthropic has withheld from public release. The model is said to have a wider use inside the Department of War, though officials have not disclosed its role. The report surfaced as Anthropic’s dispute with the Pentagon reaches the White House and as regulators expand their review of Mythos.
National Security Agency (NSA) use puts Mythos back at the center
Axios, citing two people familiar with the matter, reported that the National Security Agency is using Mythos Preview. The report said the model has gained wider use within the Department of War. The report did not explain how the National Security Agency uses Mythos in practice. It said other groups with access use the model to scan systems for exploitable security flaws.
Anthropic has kept Mythos restricted and has declined a public rollout. The company said the model can find, analyze, and explore software vulnerabilities at scale. Anthropic also said Mythos can outperform human experts in some testing environments.
During internal testing, the company said the model found thousands of critical flaws, including zero-day vulnerabilities. Cyber specialists have already warned that public access could help phishing campaigns, deepfakes, and exploit chains. Anthropic described Mythos as a “watershed moment” when it announced the model.
The Pentagon dispute reached the white house last Friday
The National Security Agency’s allegation of use comes as Anthropic’s conflict with the Pentagon has reached the White House. As previously reported , Anthropic chief executive Dario Amodei met White House chief of staff Susie Wiles.
That meeting was centered on Anthropic’s dispute with the Pentagon over limits on military use of its systems. It also follows wider testing of Mythos by federal agencies, banks, and regulators. Some administration officials see strategic value in Mythos and its cyber capabilities. One source close to negotiations said denying the government access “would be a gift to China.”
The conflict began after Anthropic refused Pentagon demands for unrestricted use of its Claude systems. The Department of War then labeled Anthropic a supply chain risk, a status usually applied to foreign companies. However, Anthropic challenged the action in court and won a preliminary injunction on March 26. The ruling paused both the directive against Anthropic technology and Pete Hegseth’s supply chain risk designation.
A later appeals ruling denied Anthropic’s request to block the designation on a temporary basis. That left Anthropic barred from Department of Defense contracts while other agency work continued. The Cybersecurity and Infrastructure Security Agency and the Treasury Department have shown interest in Mythos access. That demand increased pressure for a policy resolution inside Washington.
Australian Securities and Investments Commission (ASIC) joins other Mythos reviews
The Australian Securities and Investments Commission (ASIC) said on Monday that it is monitoring Mythos and its market implications. An ASIC spokesperson said, “ASIC is closely monitoring these developments along with peer regulators to assess possible implications for the Australian market.” The spokesperson added that ASIC works with regulators, government agencies, and the financial sector on changing technologies.
ASIC said financial services licensees should stay “on the front foot” to protect customers and clients. That statement places the Australian regulator within a wider response to Mythos. Other than ASIC, the Bank of England Governor Andrew Bailey warned that Mythos could “crack the whole cyber risk world open.” In the United States, Treasury Secretary Scott Bessent and Federal Reserve Chair Jerome Powell met other bank chiefs over Mythos.
European Central Bank President Christine Lagarde addressed the issue during an interview with Bloomberg TV. She said no governance framework currently exists “to actually mind those things.” The Bank of England’s Cross Market Operational Resilience Group and its task force also scheduled meetings on Mythos. Those sessions followed Bailey’s warning and widened the review inside British financial oversight.
Goldman Sachs chief executive David Solomon said the bank has access to the model and has “hypersensitivity” to stronger AI capabilities. JPMorgan Chase joined Project Glasswing alongside Amazon, Apple, Google, Microsoft, and Nvidia. Anthropic said it committed $100 million in credits to partners and $4 million to open-source security groups. Those reviews now place ASIC alongside United States agencies, British officials, banks, and technology firms testing Mythos.