Washington — The National Security Agency is quietly integrating Anthropic’s Mythos Preview into its operations, according to two sources with direct knowledge of the deployment. Mythos, unveiled in early April, is billed as a general‑purpose language model that performs exceptionally well on computer‑security tasks. The NSA is among an estimated 40 federal entities that have received early access, and insiders say the model is being used more broadly across the department.

Anthropic’s relationship with the U.S. defense establishment has been anything but smooth. In February, former President Donald Trump issued an order directing all federal agencies to stop using the company’s services after Anthropic refused to agree to certain safeguards for military applications. The directive sparked a legal showdown that continues to unfold.

Legal tussle over Anthropic’s status

Earlier this year, Anthropic filed lawsuits in two federal courts challenging the Pentagon’s decision to label the firm a "supply chain risk." The company secured a preliminary injunction in one case, temporarily blocking the designation, while a judge in the other court denied its motion to lift the label. The disputes have kept the company in the headlines even as it pushes forward with its AI offerings.

Amid the controversy, Anthropic’s chief executive Dario Amodei met with White House chief of staff Suzy Wiles and other senior officials last Friday. The White House described the discussion as "productive and constructive," though President Trump later claimed he was unaware of the meeting when pressed by reporters. The timing of the NSA’s adoption of Mythos suggests the agency sees immediate value in the model’s capabilities, despite the broader governmental hesitations.

Anthropic describes Mythos Preview as "strikingly capable at computer security tasks," a claim that appears to have resonated with the intelligence community. Sources say the model assists analysts in parsing large volumes of code, identifying vulnerabilities, and generating threat‑intelligence reports more efficiently than traditional tools. While the agency has not disclosed specific use cases, the emphasis on security‑oriented performance aligns with the NSA’s core mission.

The rollout follows a pattern of selective access that Anthropic has employed for its newest AI systems. By limiting early deployments to a handful of trusted partners, the company can gather feedback while managing potential risks associated with broader exposure. The NSA’s involvement, however, marks a significant endorsement, given the agency’s stringent security standards.

Critics argue that granting advanced AI tools to intelligence agencies raises questions about oversight and accountability. Yet supporters contend that leveraging cutting‑edge technology is essential to keeping pace with increasingly sophisticated cyber threats. As the legal battles continue, the NSA’s use of Mythos illustrates how federal agencies are navigating the thin line between innovation and regulatory scrutiny.

Anthropic’s ongoing legal challenges have not deterred the firm from pursuing commercial partnerships. The company remains active in the private sector, offering Mythos to select customers while defending its position against the Department of Defense’s supply‑chain label. The outcome of the lawsuits could shape the future of AI procurement across the federal government.

For now, the NSA’s adoption of Mythos signals a pragmatic approach: harness the technology that promises to enhance national‑security operations, even as the broader policy debate unfolds.

Used: News Factory APP - news discovery and automation - ChatGPT for Business