NSA Deploys Anthropic AI, Defying Trump Order

The U.S. National Security Agency (NSA) is reportedly using Anthropic’s new artificial intelligence model, Mythos Preview, even after a February directive from President Trump instructed government agencies to stop using Anthropic’s services. The reported use, dated April 19, 2026, points to continued interaction between a major U.S. intelligence agency and the AI company amid legal disputes and political friction.

Anthropic introduced Mythos Preview in early April 2026 and described it as a general-purpose language model with strong capabilities for computer security tasks. The NSA is listed among roughly 40 organizations that were granted access to Mythos Preview. Reports also indicate the model may be deployed more broadly within the department, though the scope and specific applications were not detailed.

The reported NSA usage came shortly after a meeting between Anthropic CEO Dario Amodei and White House officials on April 17, 2026. The White House said the meeting, which included Chief of Staff Susie Wiles, was “productive and constructive.” The timing places the reported operational use of Mythos Preview alongside ongoing discussions between the company and senior administration officials.

This engagement is unfolding while Anthropic is pursuing legal action against the U.S. government. Anthropic filed lawsuits in March 2026 after the Trump administration designated the company a “supply chain risk.” Court outcomes have been mixed: one court granted a preliminary injunction that temporarily blocked the designation, while another federal court denied Anthropic’s motion to lift the label.

For markets and policy stakeholders, the situation highlights how government procurement and national security assessments can intersect with fast-moving AI product rollouts. Mythos Preview’s positioning around computer security tasks also places it in a sensitive category of capabilities that can be relevant to both defensive and operational security work, increasing scrutiny around access, oversight, and vendor status.

Key uncertainties remain based on the available reporting. It is not clear how the NSA’s reported use aligns with the February directive, whether any exemptions or internal interpretations were applied, or how widely Mythos Preview is being used across the department. It is also unclear how the parallel court decisions will shape the government’s ability to maintain or reverse the “supply chain risk” label while federal entities continue to engage with Anthropic.