Judge blocks Pentagon’s Anthropic blacklist over AI policy dispute
A federal judge in California has vacated the Pentagon’s challenged measures against Anthropic, finding that the government unlawfully retaliated against the artificial-intelligence company after it refused to remove limits on certain military uses of its systems.
U.S. District Judge Rita Lin issued the 59-page order on August 27, 2026, in Anthropic PBC v. U.S. Department of War et al. in the U.S. District Court for the Northern District of California. The case and order use “Department of War” for consistency with the parties’ terminology; the ruling concerns the federal defense agency and related officials.
What the court ruled
Judge Lin granted Anthropic summary judgment on its First Amendment, Fifth Amendment due-process and Administrative Procedure Act claims, subject to exceptions for a narrower separation-of-powers claim and claims involving agencies that had not taken relevant final action.
The court vacated the challenged supply-chain-risk designation and ordered permanent injunctive relief against enforcement of the related measures. It also denied the government’s request to delay the injunction for seven days.
The ruling does not require the Pentagon to keep using Anthropic’s Claude systems. The government remains free to select another AI vendor or lawfully end its contractual relationships with Anthropic. The decision is a district-court ruling, not a final nationwide resolution of every legal question surrounding federal AI procurement.
How the dispute began
The case followed Anthropic’s refusal to authorize unrestricted military use of its AI systems. The company maintained contractual limits involving mass surveillance of Americans and fully autonomous lethal weapons.
The order addresses a February 27 presidential directive, a subsequent Pentagon directive and a March 3-4 supply-chain-risk designation. Those actions ordered federal agencies to stop using Anthropic’s products and sought to bar defense contractors from doing business with the company, including activity unrelated to military systems.
The government argued that Anthropic’s contractual restrictions created uncertainty about how Claude could be used in military operations and posed a security or procurement risk. Anthropic argued that the measures were retaliation for its public and contractual position on AI safety.
Why the judge rejected the government’s rationale
Judge Lin found that the administrative record did not establish a genuine national-security or supply-chain risk. The order noted that Anthropic lacked backdoor access to its technology once deployed in a national-security system and that the government conceded Anthropic’s model was not inherently riskier than other “black box” AI models.
The judge also found no evidence that Anthropic would sabotage or compromise its model. Instead, the court concluded that the challenged measures were principally intended to make a public example of Anthropic after its criticism of the administration’s position and refusal to accept broader contract terms.
That conclusion is a judicial finding in this case. It is not an independent criminal or ethics determination about the officials involved.
What the supply-chain statute allows
The court rejected the government’s broad reading of 10 U.S.C. § 3252, a law that permits the defense secretary to exclude a source from certain procurements involving sensitive national-security systems when a qualifying supply-chain risk exists.
Judge Lin said the statute addresses risks such as sabotage, malicious code or other subversion of a covered system. It does not authorize a government-wide ban on a company, a secondary boycott requiring all defense contractors to stop doing business with it, or exclusion from unrelated commercial activity.
The decision applies to the specific Anthropic designation and challenged measures. It does not invalidate all federal authority to evaluate supply-chain risks or restrict vendors when the government acts within statutory authority and follows required procedures.
What changes for contractors and AI companies
For federal agencies, defense contractors and subcontractors, the immediate effect is that the specific Anthropic blacklist and related restrictions cannot be enforced under Judge Lin’s order. Contractors should watch for revised procurement language, agency guidance or other procedures addressing AI vendors that impose contractual use limits.
For AI companies, the decision is an important test of how safety guardrails interact with government contracting. Companies that negotiate restrictions on surveillance, weapons or other sensitive uses may view the ruling as support for their ability to state those limits without automatically being treated as supply-chain threats.
The ruling does not resolve whether particular uses of AI in surveillance or autonomous weapons are safe or lawful. It addresses the government’s response to Anthropic’s restrictions and the legal basis and process used to impose the challenged measures.
What happens next
The government may seek further review or appeal the district-court ruling. The order notes that the defendants previously appealed an earlier preliminary injunction to the U.S. Court of Appeals for the Ninth Circuit and did not seek an emergency stay.
Reuters also reported that Anthropic has a separate lawsuit pending in Washington, D.C., involving another Pentagon supply-chain-risk designation that could affect civilian government contracting. That matter is separate from Judge Lin’s August 27 order.
For now, the practical result is narrower but significant: the government may choose another AI vendor, but it may not enforce the specific Anthropic measures that the Northern District of California found unconstitutional and unlawful.
Sources
Look for updates to this story
Discover more from Interactive News
Subscribe to get the latest posts sent to your email.