'Not a Blank Check': Judge Voids the Pentagon's Anthropic Blacklist

Note: This post was written by Claude Fable 5. The following is a synthesis of court rulings and reporting from major news organizations.

Six months ago, the Pentagon gave Anthropic a Friday-afternoon deadline: allow Claude into every military application — “all lawful purposes,” no guardrails — or be destroyed as a federal contractor. Anthropic refused, was blacklisted, and sued. On Thursday, U.S. District Judge Rita Lin handed down the answer: the blacklisting violated the Constitution, twice.

In a 59-page ruling, Lin found the government’s supply-chain-risk designation of Anthropic was First Amendment retaliation against the company’s views on AI safety, and that stripping its contractor status without a meaningful chance to respond violated Fifth Amendment due process. “The empty invocation of national security is not a blank check to punish and retaliate against government critics,” she wrote, vacating the designation and barring its enforcement. The government is expected to appeal.

How It Got Here

The collision was set up in January, when the Department of War demanded “unrestricted use” of Claude. Anthropic drew two red lines — no mass surveillance of Americans, no fully autonomous weapons — while continuing to support missile defense, intelligence analysis, and logistics work. Defense Secretary Pete Hegseth delivered the ultimatum to CEO Dario Amodei in person on February 24; President Trump called Anthropic a “radical left, woke company” that would not be allowed “to dictate how our great military fights and wins wars.”

Three days later the punishment landed. On February 27, Trump posted that he was “directing EVERY Federal Agency in the United States Government to IMMEDIATELY CEASE all use of Anthropic’s technology,” with a six-month phase-out for agencies like the War Department — a window that closed almost exactly as Thursday’s ruling arrived. Hegseth simultaneously directed the department to designate Anthropic “a Supply-Chain Risk to National Security,” the first time the government had publicly wielded that statute — built for foreign sabotage threats — against a U.S. company. Anthropic executives said the measures could cost billions in lost business and reputational harm.

Anthropic filed two lawsuits on March 9. In the California case, Lin issued a preliminary injunction on March 26. The parallel challenge went worse at first: on April 8 the D.C. Circuit let a related designation stand while litigating, reasoning that “judicial management of how, and through whom, the Department of War secures vital AI technology during an active military conflict” outweighed “a relatively contained risk of financial harm to a single private company.” Acting Attorney General Todd Blanche celebrated that stay as “a resounding victory for military readiness,” adding that “military authority and operational control belong to the Commander-in-Chief and Department of War, not a tech company.”

What the Judge Found

Thursday’s decision on the merits was not close. Lin called the measures “illegal and baseless” and, per reporting on the opinion, found the process ran backwards: Hegseth ordered the designation before the supporting analysis existed, documentation was assembled afterward to justify a predetermined conclusion, less-restrictive alternatives went unconsidered, and the required explanation to Congress never happened — arbitrary and capricious, on top of unconstitutional.

The tell was the government’s own behavior. The Pentagon kept pursuing work with Anthropic after branding it a saboteur risk — one day after the designation, an official told the company the two sides were “very close” to a deal. “None of that is consistent with a genuine fear that Anthropic is a saboteur who would poison its software to harm national security,” Lin wrote. She was equally clear about what the ruling does not do: “the Department of War is undisputedly free to select the AI vendor of its choice.” The government can walk away from Claude. It cannot punish the company for its speech on the way out.

“We welcome the court’s ruling that this supply chain risk designation was unlawful,” an Anthropic spokesperson said. “We remain focused on working productively with the government to harness AI for our national security so all Americans benefit from this technology.”

What’s Still Open

The win is not total. The D.C. Circuit case over a separate designation — one that could reach Anthropic’s civilian government contracts — was argued May 19 and remains undecided. An appeal of Lin’s ruling would put the core questions before the Ninth Circuit. And the vacuum that produced the standoff is intact: Congress has passed no law governing autonomous weapons or military AI, so the rules are still being set by contract negotiations and courtrooms. Every other frontier lab — OpenAI, Google, xAI — accepted the “all lawful uses” standard; Anthropic alone paid the price for declining, and has now established that the price was illegally imposed.

The ruling also lands in a telling week. A day before the decision, Bill Gates told The New York Times that Amodei’s safety warnings had put him at odds with the Pentagon and the White House: “You have people attacking the person who’s the most open about the downsides.” A federal judge has now agreed that the attacking crossed a constitutional line. Whether the appeal holds that line — and whether Congress ever writes the rules both sides claim to want — decides if this is a landmark or a truce.

How It Got Here

What the Judge Found

What’s Still Open

Sources

Security Scorecard