Judge Says the Pentagon’s Anthropic Blacklist Was “Illegal and Baseless”—And Violated Free Speech

A federal judge has ruled that the Pentagon crossed a constitutional line when it branded artificial-intelligence company Anthropic a national-security supply-chain risk after a public dispute over how its Claude models could be used.

In a 59-page decision issued Thursday night, August 27, U.S. District Judge Rita F. Lin found that the government’s actions amounted to unlawful retaliation under the First Amendment and denied Anthropic the process required by the Fifth Amendment. She also concluded that Defense Secretary Pete Hegseth’s designation was arbitrary and capricious under federal law.

The decision is a major legal victory for Anthropic—and a warning that invoking national security does not automatically give the government unlimited power to punish a contractor for criticizing federal policy.

The court called the measures “illegal and baseless”

Reuters reported on August 28 that Lin found no genuine factual dispute that the challenged measures violated Anthropic’s constitutional rights. The ruling said the supply-chain designation also failed to fit the statute the Pentagon invoked.

According to The Associated Press, Lin concluded that the record showed an effort to make a public example of Anthropic for criticizing the administration—not a finding that the company was likely to sabotage its technology.

The court did not order the Pentagon to purchase Anthropic’s services or continue using Claude. The ruling instead limits the government’s ability to impose broader penalties through a supply-chain-risk designation and related federal restrictions.

How an AI safety dispute became a constitutional case

The conflict began when the Pentagon sought broader authority to use commercial AI systems for military purposes. Anthropic resisted removing two central safeguards: restrictions involving mass domestic surveillance and fully autonomous weapons capable of selecting and attacking targets without meaningful human control.

Anthropic maintained that present-day AI remains too unreliable for those uses and that the restrictions were narrow enough to permit extensive national-security work. The Pentagon argued that uncertainty over when Claude might refuse a task created operational risk.

The disagreement escalated publicly. President Donald Trump and Hegseth accused the company of threatening national security, and Hegseth designated Anthropic a supply-chain risk—a label typically associated with threats to sensitive government systems and procurement.

Anthropic sued in March, arguing that the designation and the wider campaign against the company were retaliation for protected speech. Lin initially issued temporary relief. Her new ruling reaches the merits of major parts of the case and finds constitutional and administrative-law violations.

Why the First and Fifth Amendments mattered

The First Amendment generally prevents the government from punishing a person or company merely for expressing an opposing view. The Fifth Amendment requires fair procedures before the federal government deprives a party of protected interests.

Lin found both principles relevant. The court concluded that Anthropic’s public criticism was constitutionally protected and that the company was denied adequate process before receiving a designation capable of damaging its contracts, reputation and access to government business.

The decision also addressed the Administrative Procedure Act, which allows courts to overturn federal actions that are arbitrary, capricious or outside an agency’s legal authority.

The federal case docket identifies the dispute as Anthropic PBC v. U.S. Department of War in the Northern District of California. A separate, narrower Anthropic case involving another Pentagon designation remains pending in Washington.

What Anthropic and the government say

An Anthropic spokesperson welcomed the ruling and said the company still wants to work productively with the government so AI can support national security responsibly.

The Pentagon and White House had not issued a detailed public response when AP and Reuters published their reports. AP said the government is expected to continue fighting the decision.

Technology trade groups also welcomed the result. The Computer & Communications Industry Association, which participated in supporting briefs, said the case matters to any company doing business with the federal government because national-security procurement tools can carry consequences far beyond one contract.

The bigger fight over AI and warfare is not over

The ruling settles neither the technical question of how reliable advanced AI is in combat nor the policy question of how much control a private company should retain after selling technology to the military.

It does, however, establish an important boundary: the government may choose its vendors, but it cannot automatically convert a contract dispute and public criticism into sweeping punishment without evidence, lawful authority and due process.

That distinction will matter as the Pentagon increasingly relies on privately developed AI for intelligence analysis, planning, logistics and battlefield systems. The most important rules may emerge not only from Congress or military policy—but from courts deciding where national-security power ends and constitutional protections begin.