Judge rules Pentagon unlawfully retaliated against Anthropic

A federal judge on Thursday ruled that the administration unlawfully retaliated against Anthropic after the artificial-intelligence company refused to remove contractual limits on mass surveillance of Americans and lethal autonomous warfare.

U.S. District Judge Rita Lin of the Northern District of California found that the government violated Anthropic’s First Amendment rights, denied the company required Fifth Amendment due process and acted unlawfully under the Administrative Procedure Act.

The order, filed August 27, 2026, says the Pentagon’s supply-chain-risk designation was contrary to the governing statutory scheme and arbitrary and capricious. It also says the record did not support the government’s stated national-security rationale.

The court granted Anthropic summary judgment on the central claims, but not on every claim. It rejected the company’s separate ultra vires separation-of-powers challenge and entered judgment for the government concerning certain agencies that did not take relevant action or took only interim measures.

What the judge found

The dispute began as the Pentagon sought an “all lawful uses” provision for Claude, Anthropic’s AI model. Anthropic agreed to remove most restrictions but kept limits on two uses: mass surveillance of Americans and lethal autonomous warfare.

Anthropic said those uses raised safety and civil-rights concerns. The Pentagon argued that private companies should not be able to constrain military operations and later designated Anthropic a national-security supply-chain risk.

Judge Lin concluded that the government’s administrative record was thin. The ruling says a four-page memorandum supplied most of the stated justification and did not establish that Anthropic could sabotage or alter its models after deployment. The court also pointed to the government’s prior approval and use of Claude, along with continued efforts to work with Anthropic after the designation.

Those facts, the judge concluded, were inconsistent with the administration’s claim that Anthropic posed a genuine sabotage threat. The court found that the challenged actions were tied instead to Anthropic’s public criticism and its refusal to accept the requested contract language.

“The empty invocation of national security is not a blank check to punish and retaliate against government critics,” Lin wrote in the order.

What changes now

The court said Anthropic is entitled to vacatur and permanent injunctive relief, but it also said that the operative relief would be issued in a separate order. The judge denied the government’s request to administratively stay the permanent injunction for seven days, noting that officials had already complied with the preliminary injunction for more than five months.

Until the separate relief order and any appellate action are clear, agencies and defense contractors will need to examine the precise scope of the ruling and how it affects existing restrictions, contracts and procurement systems.

The decision does not require the Pentagon or any other agency to use Claude, restore Anthropic to every federal contract or select the company as a vendor. The court expressly recognized the government’s authority to choose the AI products that meet its operational needs through lawful procurement procedures.

That distinction is central. The case was not about whether the government must continue buying Anthropic’s technology. It was about whether officials could impose sweeping government-wide and contractor restrictions as punishment for the company’s position on high-risk AI uses and its public criticism of the administration.

Two legal tracks remain

The administration is expected to appeal, but an appeal should not be described as filed unless a notice of appeal or appellate motion is confirmed. The government could also seek an emergency stay in the appeals court.

A separate case remains pending in the U.S. Court of Appeals for the D.C. Circuit. That litigation concerns a March 3, 2026 designation under a different statute, 41 U.S.C. § 4713. In an April order, the appeals court denied Anthropic’s request for a stay but granted expedited consideration, leaving the merits unresolved.

The two cases therefore involve related events but different legal tracks. The Northern District of California ruling addresses the broader government-wide restrictions and the designation challenged under 10 U.S.C. § 3252. The D.C. Circuit matter concerns the separate Section 4713 action barring Anthropic from providing goods or services to the Defense Department.

Why the ruling matters

For AI companies, defense contractors and federal procurement officials, the decision is a significant test of how far national-security and supply-chain authorities may reach when a vendor disputes the government over surveillance, autonomous weapons or other sensitive AI applications.

The ruling does not eliminate the government’s ability to reject a vendor for legitimate operational, security or business reasons. It says broad punitive restrictions cannot be justified by an unsupported national-security rationale or imposed in retaliation for protected speech.

The next major markers are the court’s separate relief order, any emergency stay request, a possible government appeal and further proceedings in the D.C. Circuit. Those actions will determine how quickly the ruling changes procurement practice and how much discretion future administrations retain when technology companies challenge high-risk uses of artificial intelligence.

Sources

Look for updates to this story

Discover more from Interactive News

Subscribe to get the latest posts sent to your email.