Judge rules Pentagon Anthropic blacklist unlawful

A federal judge has ruled that the Pentagon's designation of Anthropic as a supply-chain risk was unlawful, granting the company summary judgment in part on First Amendment and administrative-law grounds. The order, filed August 27 by Judge Rita F. Lin in the Northern District of California, is a win, but not a full clearance: a separate order on remedy is still coming, and a parallel case in Washington, D.C., is still open.

What the court actually decided

This is a ruling on cross-motions for summary judgment, not a jury verdict or a settlement. Judge Lin granted judgment for Anthropic in part on three theories: First Amendment retaliation (Count II), due process under the Fifth Amendment (Count IV), and Administrative Procedure Act challenges to the Hegseth Directive and the Supply Chain Designation under 10 U.S.C. § 3252 (Count I), as to Defense Secretary Pete Hegseth and the Department of War. She denied Anthropic's ultra vires claim, so the company did not win on every count.

The language is unusually direct for a designation the government defended on national-security grounds. Lin wrote that the measures were "illegal and baseless," that they reflected "a desire to make a public example out of Anthropic for its 'arrogance' in criticizing the government," and that "the empty invocation of national security is not a blank check to punish and retaliate against government critics." The court also found the company "was denied the pre-deprivation process required under the Fifth Amendment." The administrative record the government relied on was thin, a short memo, which is what made the APA arbitrary-and-capricious finding available.

The step the government did not get

Here is the procedural detail the one-line headlines skip. The court refused to give the government a running start. Defendants asked to stay the forthcoming permanent injunction for seven days; Judge Lin denied that request, and noted that "an order addressing relief will issue separately." So the remedy, likely including an injunction against the designation, is being written now, and the government does not get a week's grace before it lands.

Why this is a win, not a clearance

Do not read this as Anthropic being restored to all Defense Department work. The dispute ran on two tracks: Anthropic filed complaints in both California and D.C. in March, and the D.C. suit is still ongoing. Related federal supply-chain-security mechanisms could still leave the company constrained on DoD contracting until that case resolves, regardless of today's California result. This ruling strikes down the specific Hegseth designation as unlawful retaliation; it does not, by itself, guarantee the contracts flow again.

The takeaway

The precedent is the asset here, not the immediate contract access. For anyone selling AI to the government, the operative finding is that a supply-chain-risk label imposed to punish a vendor's public criticism is First Amendment retaliation a court will strike, and that "national security" alone will not shield it. Watch three things next: the separate relief order for how broad the injunction runs, whether the government appeals inside its one-week window, and the D.C. case, which is the one that still decides whether Anthropic is actually clear to contract. Until the relief order and the D.C. track resolve, treat this as a strong ruling on principle that has not yet reopened the door.

For related context on Anthropic's trajectory, see our coverage of the OpenAI data-center leadership exit reshaping the same buildout race, and on Anthropic's own scale, its $45B Nscale compute lease and $30T TAM IPO pitch.