The National Security Agency is reportedly using Anthropic’s Mythos Preview model to scan network environments for vulnerabilities, according to Axios. The development follows a Department of Defense assessment that previously classified Anthropic as a potential supply-chain risk after the company limited access to certain model capabilities.

NSA Integration And Cybersecurity Applications

Anthropic introduced Mythos as a model tailored for cybersecurity applications, with restricted availability due to concerns over misuse in offensive cyber operations. Access has been granted to a limited group of around 40 organizations, with only a subset publicly identified. The controlled rollout reflects an effort to balance capability with risk management.

Follow THE FUTURE on LinkedIn, Facebook, Instagram, X and Telegram

Balancing National Security And Operational Risks

Sources indicate that the NSA is among the organizations using Mythos, primarily to detect vulnerabilities across network systems. The U.K.’s AI Security Institute has also confirmed access, pointing to broader institutional interest in the model’s capabilities.

Disputes With The Pentagon

Use of Anthropic’s tools by U.S. agencies comes amid ongoing tensions with the Pentagon. Legal arguments have raised concerns that such technologies could pose national security risks. The dispute intensified after Anthropic declined to provide full access to its Claude model for applications related to mass surveillance and autonomous weapons.

Thawing Relations With Government Leadership

Recent engagement suggests a shift in tone between Anthropic and U.S. officials. Anthropic CEO Dario Amodei recently met with White House Chief of Staff Susie Wiles and Treasury Secretary Scott Bessent, with officials describing the discussions as productive. Requests for comment have been sent to the NSA, while Anthropic has not issued additional statements.