A U.S. judge on Thursday blocked the Pentagon from blacklisting Anthropic as a national security supply-chain risk, handing the Claude maker a significant legal victory in its escalating dispute with the Trump administration over how artificial intelligence should be used in military operations.

U.S. District Judge Rita Lin ruled that Defense Secretary Pete Hegseth had exceeded his authority when he designated Anthropic a supply-chain risk, a move that restricted the company from certain military contracts and threatened to cut off a potentially lucrative government market.

In a 59-page order, Lin described the Pentagon’s decision as “illegal and baseless,” finding that the government could not invoke national security as a justification for retaliating against a company over its position on AI safety.

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

The ruling marks a major setback for the Pentagon’s attempt to force AI companies to accept broader military uses of their systems. It also creates an important legal test for the government’s authority to penalize technology companies whose safety policies conflict with military requirements.

The dispute began after Anthropic refused to allow its Claude models to be used for certain forms of U.S. surveillance and autonomous weapons. The company has argued that current AI systems are not sufficiently reliable for autonomous weapons and has raised concerns about domestic surveillance and individual rights.

The Pentagon has taken a fundamentally different position, arguing that private technology companies should not be able to impose restrictions on how the U.S. military conducts lawful operations.

Hegseth’s decision was unprecedented. It marked the first public designation of a U.S. company as a supply-chain risk under a relatively obscure government procurement authority designed to protect military systems from potential infiltration or sabotage by adversaries.

The consequences extended well beyond the immediate contracts of Anthropic affected by the designation. Company executives had warned that the decision could cost billions of dollars in business and damage its reputation among government and commercial customers.

Anthropic welcomed Thursday’s ruling.

“We remain focused on working productively with the government to harness AI for our national security so all Americans benefit from this technology,” the company said.

A Broader Fight Over Who Controls Military AI

Anthropic’s legal challenge goes to the heart of a rapidly emerging question in the AI industry: how much control should model developers retain over the uses of increasingly powerful systems once those systems are supplied to governments?

The company believes that restrictions on autonomous weapons and domestic surveillance are safety and civil-liberties positions rather than an attempt to obstruct national security operations. Its lawsuit filed in March accused the government of violating its First Amendment rights by retaliating against its position on AI safety.

Anthropic also argued that the Pentagon violated its Fifth Amendment right to due process because it was not given an opportunity to challenge the supply-chain-risk designation before it was imposed.

The Justice Department has rejected that characterization. According to court filings, the government said that Anthropic’s refusal to remove its restrictions could create uncertainty about how the Pentagon could deploy Claude and potentially leave military systems without access to critical capabilities during operations.

The government maintained that the designation was triggered by Anthropic’s refusal to accept contractual conditions, rather than by the company’s broader views on AI safety.

That could prove important as the administration seeks to expand the use of AI across the U.S. military. Currently, the government is largely dependent on private AI developers for models and computing infrastructure, giving companies such as Anthropic, OpenAI and other contractors greater influence over how those technologies can be deployed.

The case therefore pits two competing principles against each other. The Pentagon wants assurances that AI systems purchased for national security can be used when and where military commanders require them. Anthropic argues that developers must retain limits on applications they consider unsafe or incompatible with fundamental rights.

Anthropic Faces Another Government-Contract Battle

Thursday’s ruling does not resolve Anthropic’s broader confrontation with the U.S. government. The company has a separate lawsuit pending in federal court in Washington, D.C., challenging another Pentagon supply-chain-risk designation. That case could have consequences for Anthropic’s eligibility for civilian government contracts, potentially widening the dispute beyond military procurement.

The California ruling could nevertheless strengthen Anthropic’s position in that litigation and provide technology companies with a significant precedent when challenging government procurement decisions that they believe punish them for corporate policies or public positions.

The outcome is also likely to be closely watched by other AI developers as they negotiate contracts with the Pentagon and other national-security agencies. The commercial stakes are substantial as the U.S. government accelerates investment in AI and seeks access to the most capable models.

However, the ruling leaves unresolved the underlying policy disagreement over autonomous weapons and surveillance. But it establishes an important limit, at least for now, on how the government can use its procurement powers against an AI company that refuses to accept certain military applications of its technology.