The Pentagon Wins a Major Round Against Anthropic
The Pentagon has scored a significant victory in its dispute with Anthropic. A US appeals court now allows the military to treat the maker of Claude as a supply chain risk.
The designation follows the company’s refusal to lift its limits on AI use. Those limits cover fully autonomous weapons and mass surveillance.
A panel of the D.C. Circuit backed the Pentagon by a vote of two to one. You can read the full text in the published court decision.
How the Dispute Began
The conflict emerged after months of negotiations. Throughout them, military officials demanded that Claude be usable “for any lawful purpose.”
Anthropic agreed to widen access substantially. Nevertheless, it kept two restrictions in place.
The Two Red Lines
First, no one may use the model for mass surveillance of Americans. Second, no one may use it in weapons that select and strike targets without human involvement.
Why the Majority Sided With the Pentagon
Most judges concluded that the Pentagon acted within the law. In their view, Claude’s built-in limits could disrupt military systems at a critical moment.
The case files mention several incidents. In some, the model refused authorized government tasks. In addition, a separate dispute arose over Claude’s use during a classified overseas operation.
The Free Speech Argument Fails
The court also rejected Anthropic’s central claim. The company argued that officials punished it for its public stance on AI safety.
The majority disagreed. It saw the cause of the conflict not in the developer’s views, but in its refusal of a contract term the Pentagon needs.
A Dissent on the Statute
One of the three judges dissented. He argued that the supply chain risk law should not reach openly stated product limits. Moreover, he noted the absence of any malicious intent.
A Legal Fight That Is Far From Over
This ruling does not settle every lawsuit. In August, a federal court in California reached the opposite conclusion in a parallel case. It declared broader measures against Anthropic unlawful.
Judge Rita Lin blocked a government-wide ban on the company’s products. She also blocked a demand that military contractors cut all ties with it.
The Washington panel, however, examined a different legal basis. Its decision states that the Pentagon may remove Claude from its supply chain. After all, the department need not risk an AI refusal during combat operations.
Meanwhile, the military has spent months seeking alternatives to Claude. It is also expanding work with other AI developers.
Anthropic’s Next Steps
Anthropic says it disagrees with the panel’s conclusions. The company is weighing further appeals, including a rehearing before the full appeals court.
It claims the risk label has already cost it billions of dollars. Furthermore, the company says the label has damaged its reputation.
Washington Still Relies on Anthropic Technology
Dropping Anthropic entirely has proved hard for US security agencies. In August, reports showed that the NSA kept testing Mythos. This happened even as the Pentagon tried to push the company’s products out of military infrastructure.
Claude’s Growing Power Raises New Fears
At the same time, Claude’s abilities edge closer to a scenario Anthropic itself fears. The company recently described operations in which a human merely chose the target. The AI then handled reconnaissance, vulnerability exploitation, and other stages of a cyberattack.
Surveillance forms a separate front. Anthropic’s research suggests that AI can slash the cost of processing vast troves of personal data. As a result, state surveillance may need far fewer analysts.
Support Our Threat Intelligence
If you find our technology report and cybersecurity news helpful, consider supporting our work.
Buy Me a Coffee
PayPal
