Pentagon vs. Anthropic: a district court vacates one designation, the D.C. Circuit upholds the exclusion

The Pentagon asked Anthropic to allow its Claude AI models for "all lawful uses." Anthropic refused to drop two contract limits: no use for "lethal autonomous warfare" and no "mass surveillance of Americans." On Feb 27, 2026, President Trump directed every federal agency to stop using Anthropic's technology, and Secretary of War Hegseth directed the Department to designate Anthropic a supply chain risk. Hegseth made the formal determinations on March 3. A March 6 Department of War memo ordered Claude removed from its systems and barred contractors from using it in their work for the Department. Anthropic challenged the actions in two courts, under two different laws.

On Aug 27, Judge Rita F. Lin (N.D. Cal., No. 26-cv-01996) largely ruled for Anthropic in a 59-page opinion, finding the broad measures "illegal and baseless." She vacated the "supply chain risk" designation under 10 U.S.C. § 3252 and the order barring military contractors from any commercial activity with Anthropic, and declared the actions First Amendment retaliation and a denial of due process. The government won on Anthropic's separation-of-powers claim and on claims against some agencies.

On Sept 25, the D.C. Circuit, which Congress gave exclusive jurisdiction to review procurement actions under the Federal Acquisition Supply Chain Security Act (41 U.S.C. § 4713), denied Anthropic's petitions 2-1 (Anthropic PBC v. U.S. Department of War, No. 26-1049). Judge Katsas, joined by Judge Rao, pointed to a record in which Claude's built-in restrictions had "stopped Claude from performing tasks requested by government users," and to a dispute over whether Anthropic's contract barred Claude's use "in an ongoing overseas military operation." The court held that the Department "reasonably feared that Anthropic might manipulate Claude's design to prevent it from performing national-security functions" it considers authorized. It said it had "no quarrel" with Judge Lin's finding that Anthropic acted with no bad motive, but held that section 4713, unlike section 3252, does not require one. It also rejected Anthropic's First Amendment claim, finding the exclusion turned on a contract term, not on Anthropic's advocacy.

Judge Henderson dissented, writing that under the majority's reading a contractor faces a choice: "Agree to the Secretary's demands or risk being designated a national security threat under FASCSA."

How we know: Court records read directly: the D.C. Circuit opinion and dissent, and Judge Lin's order, relief order and judgment. Not yet confirmed: whether either side has sought further review, and what happens to the government's Ninth Circuit appeal of the March 26 preliminary injunction (No. 26-2011), which was on hold pending the D.C. Circuit's ruling. Updated Oct 1, 2026; earlier versions ended with the Aug 27 ruling. Conflict-control review (see Method): drafted with an Anthropic model, then checked against the court records by OpenAI's GPT-5.6 Sol, including for wording that runs one way. OpenAI has its own stake: the D.C. Circuit notes the Department expanded its relationship with OpenAI after the exclusion.