DoD Accused of Punishing Anthropic for AI Military Use Restrictions | WIRED

A U.S. District Judge suggested the Pentagon may be illegally retaliating against Anthropic, a leading artificial intelligence company, for attempting to restrict the military’s use of its AI tools. Judge Rita Lin, during a hearing Tuesday in San Francisco, described the Department of Defense’s actions as appearing to be “an attempt to cripple Anthropic,” according to court transcripts.

The dispute centers on Anthropic’s efforts to limit how its AI model, Claude, could be deployed by the armed forces, specifically preventing its use in fully autonomous weapons systems or for surveillance of American citizens. The Pentagon responded by designating Anthropic a “supply chain risk,” a move that effectively discourages government contractors from using the company’s technology. Anthropic has filed two federal lawsuits challenging the designation, alleging it amounts to unlawful retaliation by the Trump administration.

During the hearing, Judge Lin questioned the government’s rationale for the designation, stating that the actions “don’t seem to be tailored to the stated national security concerns.” She pressed Eric Hamilton, a Justice Department attorney representing the Pentagon, on whether less punitive measures had been considered. The supply chain risk designation, Lin noted, is typically reserved for hostile actors like foreign adversaries and terrorists.

Hamilton argued that the Department of Defense, now referred to internally as the Department of War (DoW), had legitimate concerns about Anthropic potentially manipulating its AI models to undermine military operations. “The worry is that Anthropic, instead of merely raising concerns and pushing back, will say we have a problem with what DoW is doing and will manipulate the software… so it doesn’t operate in the way DoW expects and wants it to,” he said.

However, Judge Lin challenged this assertion, pointing out that the DoW could simply cease using Claude if its functionality was a concern. She also questioned why Defense Secretary Pete Hegseth issued a public directive on X (formerly Twitter) last month, stating that no military contractor could conduct commercial activity with Anthropic. Hamilton conceded under questioning that Hegseth lacked the legal authority to broadly bar contractors from using Anthropic for non-military related work, but stated he did not grasp why the Secretary made the public statement.

Anthropic is seeking a temporary order to pause the supply chain risk designation, hoping it will reassure customers who have become “skittish” in light of the dispute. A ruling on the injunction is expected within the coming days. The company argues the designation has jeopardized its business prospects.

The Pentagon has indicated it is working to replace Anthropic’s technology with alternatives from Google, OpenAI, and xAI, and has asserted it has measures in place to prevent any tampering with its systems during the transition. Hamilton stated he was unsure if Anthropic could even update its AI models without Pentagon permission, a claim the company disputes.

A separate case filed by Anthropic in the federal appeals court in Washington, D.C. Is also pending, with a ruling expected soon without a hearing.