Judge Rules Pentagon Anthropic Blacklist Was Illegal Retaliation

Judge Rita Lin's order Thursday night is the first judicial precedent blocking a supply-chain-risk designation under 10 U.S.C. § 4872 on First Amendment grounds, a statute that, until now, had been used against entities tied to foreign intelligence services like Huawei, ZTE, and Kaspersky. The statute contains no hearing requirement and no built-in judicial review. Lin found that the government stretched it past its breaking point.

The sequence is well-established in the record. In late February 2026, contract negotiations between Anthropic and the Pentagon collapsed after Defense Secretary Pete Hegseth demanded unrestricted military use of Claude. Anthropic sought narrow carve-outs for fully autonomous weapons and mass surveillance of Americans. On February 28, Trump and Hegseth threatened the supply-chain designation. On March 6, the Pentagon made it official, "effective immediately." Hours later, OpenAI announced its own Pentagon deal with no use restrictions. Anthropic sued March 9. Lin issued a temporary restraining order March 26. The D.C. Circuit declined parallel relief in April, creating a split the government flagged in subsequent filings.

The ruling drew a bright line the government's lawyers had argued didn't exist. "Nothing in the governing statute supports the Orwellian notion that an American company may be branded a potential adversary and saboteur of the U.S. for expressing disagreement with the government," Lin wrote. At a July 30 hearing, she said the record had "gotten worse for the government" over time and called the government's position "at odds with the First Amendment."

What the government argued, and what Lin rejected

DOJ lawyers contended that AI models are "so staggeringly enormous and opaque" that the Defense Department cannot evaluate them like physical hardware. That argument (that the opacity of frontier models justifies an effectively unreviewable designation) is the closest thing to a limiting principle the government offered. Lin didn't buy it. Her ruling treats the designation as a penalty imposed for speech, not a risk assessment grounded in evidence of sabotage or compromise.

The D.C. Circuit's May hearing previewed the government's likely appeal posture. Judge Neomi Rao, a Trump appointee, questioned what basis a court has for second-guessing the Secretary of Defense's judgment about risk. Judge Karen LeCraft Henderson, a George H.W. Bush appointee, called the Pentagon's action "a spectacular overreach." That division is exactly what the government needs to keep the case alive.

What changes Monday

Federal contractors and agencies that stopped using Claude after the designation can resume. The ruling vacates the label and the Trump social-media directive ordering agencies to drop Anthropic's products. For the broader defense-industrial base, the precedent is the larger prize: a vendor can now challenge a supply-chain designation in court and win, which changes the risk calculus for any AI company weighing public policy advocacy against future contract opportunities.

The D.C. Circuit case remains pending. DOJ is expected to appeal Lin's ruling to the Ninth Circuit. A circuit split would tee up the question for the Supreme Court, and the answer would determine whether every defense contractor's First Amendment rights stop at the Pentagon's procurement door.