What happened

Anthropic, the AI company that makes Claude, lost a major legal challenge on September 25. The U.S. Court of Appeals for the D.C. Circuit upheld the Pentagon’s decision to exclude Claude from its supply chain. The court denied Anthropic’s petitions for review. The court opinion says the government acted under the 2018 Federal Acquisition Supply Chain Security Act.

The ruling was reported by CNBC. The story also drew attention on Hacker News, a technology discussion site. Its post received 370 points and 682 comments. Those numbers show community attention. They do not prove that the ruling or the report is correct.

Why there was a fight

Anthropic allows some military work with Claude. It keeps two major limits. It does not allow fully autonomous lethal warfare. It also rejects mass surveillance of people in the United States. The company refused a Pentagon demand for contract terms allowing ‘all lawful uses’ of Claude.

The Pentagon argued that these limits could create a supply-chain risk. Officials worried that Claude might not work when a military task needed it. They also worried that Anthropic could block a use through its contract or technical controls. The court said this risk could fit the broad procurement law. It said the government did not need to prove malicious intent under this law.

An important detail is the existence of another case. In August, a California federal court set aside a separate Anthropic designation. That case used a different statute. The statute had a narrower definition. It focused more on adversaries and malicious interference. The D.C. Circuit said the two cases involved different legal questions. The same phrase, ‘supply-chain risk,’ did not mean the same legal test in both cases.

Why it matters

The ruling is not a technical finding that Anthropic is malicious. It is not a test proving that Claude is unsafe. The court mainly asked whether the government’s procurement decision was reasonable under the law. Because national security was involved, the court gave the executive branch substantial room to judge the risk.

That creates a difficult policy question for the AI industry. Safety limits can protect people. Yet a government customer may see those limits as a loss of control. This ruling gives the government more room to exclude a supplier when it believes the supplier may restrict needed uses. The court did not decide which general policy is always best. It accepted the government’s authority to choose in this case.

What the court actually decided

The majority upheld the exclusion of Claude from the Pentagon’s supply chain. It also rejected Anthropic’s due-process and free-speech arguments. The court found that any problem with the timing of notice had not harmed Anthropic’s case.

Judge Karen LeCraft Henderson dissented. She argued that the law targeted covert sabotage, hostile actors, and malicious manipulation. In her view, openly enforcing safety restrictions should not qualify as a supply-chain risk. The direct ruling concerns the Pentagon’s procurement system. It is not an automatic ban on Claude for every federal agency, company, or private user.

What remains unknown

The candidate report and the opinion do not show whether Anthropic will seek review by a higher court. They also do not show how broadly the Pentagon will enforce the exclusion. The effect on contractors, government agencies, and Anthropic’s business remains uncertain. The court did not evaluate the real-world safety of autonomous weapons or mass surveillance systems.

What to watch next

Watch Anthropic’s next legal move and the Pentagon’s implementation. Watch whether contractors move from Claude to other models. Other AI companies may also rethink their government contracts and usage limits. Congress could eventually clarify how broadly the supply-chain law should apply. The Hacker News numbers show that the case matters to a technology community. They remain a measure of attention, not a verdict on the facts.