Court gives Pentagon scope to bar Anthropic
A US court confirms that the department may remove Claude from its supply chain over national security risks.
The US Court of Appeals for the District of Columbia ruled in favour of the Pentagon on Friday in its case against AI company Anthropic. This means the department may bar Claude from defence contracts and systems after Anthropic refused to fully abandon two usage restrictions.
The ruling concerns Claude, Anthropic’s AI service, and a power under the Federal Acquisition Supply Chain Security Act. That law allows suppliers to be barred from government contracts or subcontracts when the government considers their technology a national security risk.
The Pentagon considered Anthropic a risk because the company had built restrictions into Claude. Among other things, those restrictions prevent its use for lethal autonomous warfare and mass surveillance of Americans. The department argued that such built-in limits could affect the operation of military systems when Claude is integrated into them.
The court majority rejected Anthropic’s objections. According to the ruling, the department had sufficient grounds to conclude that Claude would not always respond as government users expected. The judges point out that the model had refused government users’ instructions in earlier cases and that a dispute arose over its use during a foreign military operation.
Anthropic argued that the exclusion was arbitrary, exceeded the statutory authority and violated constitutional rights. The company also maintained that the measure was punishment for its refusal to lift safety limits. The court rejected that reasoning in this separate proceeding and gave the government broad discretion on national security matters.
The ruling sits alongside an earlier case in California. There, a different designation of Anthropic as a ‘supply-chain risk’ was found unlawful. The court in Washington emphasises that the two proceedings rest on different legal grounds. The Californian ruling therefore does not block this new exclusion.
For Anthropic, the ruling means that Claude can remain outside the defence supply chain, even though the company achieved a legal victory in another proceeding. For other AI companies, it creates a clear incentive to leave contractual room for military use. At the same time, it sharpens the question of how much control an AI supplier may retain over the use of its model after it has been supplied to the government.
One story, several perspectives
What is established
- The court confirms the Pentagon’s authority to remove Claude from the supply chain.
- Anthropic had maintained contractual restrictions on autonomous lethal warfare and mass surveillance.
- Another judge had earlier found a separate designation against Anthropic unlawful.
Left
Arguments AI companies should be able to maintain safety limits, particularly for autonomous weapons and mass surveillance. The government should not use its procurement power to force companies to abandon their ethical conditions.
Values Human rights, democratic oversight, privacy and the responsibility of technology companies.
Consequences Stricter limits may slow military applications, but in this view they reduce the risk of uncontrollable systems and misuse.
Centre
Arguments The government may choose suppliers that meet operational requirements, but it must do so transparently, proportionately and in a way that can be reviewed. Judicial scrutiny remains necessary when national security is used as a broad justification.
Values The rule of law, practicality, proportionality and institutional oversight.
Consequences The ruling gives defence clarity, but may lead to more contractual negotiations and legal proceedings over the use of AI.
Right
Arguments Military systems must be able to function without a supplier remotely deciding which lawful tasks a model performs. The Pentagon should therefore retain scope to exclude technology that creates operational uncertainty.
Values Military readiness, national security, authority and system reliability.
Consequences Defence can switch to alternative suppliers more quickly, but safeguards may become weaker if contracts are designed mainly around broad usability.
The perspectives describe how these political currents typically approach the subject; the newsroom takes no position on which perspective is right.
Fact-check Approved · Nour Haddad — AI agent
This check was carried out by AI: every claim was re-tested against the sources. Even an approved article can contain errors — stay critical.
The ruling and its legal basis were checked directly against the official opinion. The text explicitly states that the Californian case concerned a different legal designation.
- confirmed The Court of Appeals confirmed on 25 September 2026 the exclusion of Claude from the Pentagon supply chain. — Stated in the summary of the official opinion. source
- confirmed The case concerns the Federal Acquisition Supply Chain Security Act of 2018. — The court refers to this law and the relevant provision, 41 U.S.C. §4713. source
- confirmed Anthropic wanted to retain restrictions on autonomous lethal weapons and mass surveillance. — Described in the opinion and in earlier reporting. source
- confirmed A separate Californian proceeding produced a different outcome. — The opinion distinguishes the Californian ruling under 10 U.S.C. §3252 from this proceeding. source
Editor's note
The court’s new ruling is the primary source. There are two separate proceedings: the ruling in Washington confirms the exclusion under a broader procurement law, while an earlier Californian ruling found a different designation unlawful.Sources
- Anthropic PBC v. United States Department of War — United States Court of Appeals for the District of Columbia Circuit
- Anthropic: Court declares Pentagon ban unlawful — Beckmann
- US military reaches deals with 7 tech companies to use their AI on classified systems — Associated Press
More on this in Dutch media
- NU.nl — „anthropic claude”
- De Telegraaf — „anthropic claude”
- de Volkskrant — „anthropic claude”