Two months after President Trump ordered every federal agency to stop using Anthropic's products, the National Security Agency is already running the company's newest model. Mythos Preview — a general-purpose system Anthropic says is "strikingly capable at computer security tasks" — landed inside the NSA as part of a broader preview rollout to roughly 40 organizations, according to Axios, which cited two sources familiar with the arrangement. One of those sources said the model is "being used more widely within the department" beyond just the NSA.
The quiet adoption is remarkable given how publicly the relationship between Anthropic and the Trump administration has fractured. In February, the White House blacklisted the AI company after it refused to weaken certain safety guardrails for military applications during contract negotiations. The Pentagon later branded Anthropic a "supply chain risk" — a designation the company has fought in two separate federal courts. It won a preliminary injunction in one. It lost in the other.
And yet, on Friday, Anthropic CEO Dario Amodei walked into a meeting with White House Chief of Staff Susie Wiles and Treasury Secretary Scott Bessent.
"We discussed opportunities for collaboration, as well as shared approaches and protocols to address the challenges associated with scaling this technology," the White House said afterward, calling the sit-down "productive and constructive." Anthropic's own statement echoed the tone, describing "a productive discussion on how Anthropic and the U.S. government can work together on key shared priorities such as cybersecurity, America's lead in the AI race, and AI safety." The company said it's "looking forward to continuing these discussions."
Trump himself appeared unaware of the meeting. Asked by reporters, he said he had "no idea" it had taken place.
The gap between the official ban and the NSA's actual use of Mythos suggests the administration's posture toward Anthropic is neither uniform nor settled. There were earlier signals: reports indicated that Bessent and Federal Reserve Chair Jerome Powell had been encouraging heads of major banks to test Mythos as well. Anthropic co-founder Jack Clark leaned into this framing, calling the supply-chain risk designation a "narrow contracting dispute" that wouldn't stop the company from briefing the government on its latest models. In other words, the legal fight and the technical relationship appear to be running on parallel tracks.
That legal fight, though, isn't going away. Anthropic filed lawsuits against the Department of Defense in two courts in March, challenging the supply-chain label that effectively locks the company out of lucrative government contracts. The Pentagon filed responses shortly after. One court sided with Anthropic on a temporary basis; the other did not. No trial dates have been set.
For now, the most sensitive intelligence agency in the country is testing an AI model built by a company the Defense Department has officially deemed a risk — while that company's CEO shakes hands at the White House. The next round of court filings could determine whether the arrangement lasts.
