DoW Appears to Illegally Retaliate Against Anthropic for AI Military Use Limits, Federal Judge Says

During a Tuesday court hearing, U.S. District Judge Rita Lin has concluded that the U.S. Department of Defense—recently rebranded as the Department of War (DoW)—appears to be illegally punishing AI developer Anthropic for imposing restrictions on military use of its artificial intelligence tools.

Describing the Pentagon’s decision to label Anthropic a national supply chain security risk, Lin told the court: “This looks like an intentional effort to cripple Anthropic. It appears the department is penalizing the company for drawing public scrutiny to this contract dispute, which would clearly count as a violation of the First Amendment.”

Anthropic has filed two separate federal lawsuits arguing that the Trump administration’s security risk designation amounts to unlawful retaliation. The government applied the controversial label immediately after Anthropic pushed to formalize limits on how its AI technology could be utilized by the U.S. military. Tuesday’s hearing was held for the first of the two cases, which was filed in San Francisco.

The AI firm is seeking a temporary court order to suspend the harmful designation. Anthropic says this relief is critical to convincing its already skittish, on-the-fence customers to continue partnering with the company while the full legal process plays out. Lin can only grant the suspension if she determines that Anthropic is likely to prevail on the merits of its overall lawsuit, and a ruling on the requested injunction is expected within the next several days.

The high-stakes conflict has sparked a much broader public debate around two pressing issues: the growing integration of artificial intelligence into U.S. military operations, and whether Silicon Valley tech companies should defer fully to the U.S. government when deciding how the technologies they build are deployed.

The DoW has defended its actions, arguing it followed all formal protocols and properly concluded that Anthropic’s AI tools can no longer be trusted to perform as required during critical operations. The administration has urged Lin not to second-guess its assessment of the national security threat it claims Anthropic poses.

During Tuesday’s hearing, Trump administration attorney Eric Hamilton laid out the government’s core concern: “The worry is that Anthropic, instead of merely raising concerns and pushing back, will object to what the DoW is doing and manipulate its own software … so it does not operate in the way the DoW expects and wants.”

Lin acknowledged that deciding whether Anthropic is an appropriate government vendor falls to Defense Secretary Pete Hegseth, not the judiciary. However, she added, it is explicitly her role to determine whether Hegseth broke the law by taking punitive action far beyond simply canceling Anthropic’s existing government contracts. Lin said she found it “troubling” that both the supply chain risk designation and broader directives restricting government contractors’ access to Anthropic’s flagship AI Claude “don’t seem to be tailored to stated national security concerns.”

As the dispute between Anthropic and the administration escalated last month, Hegseth posted a sweeping announcement to social platform X: “effective immediately, no contractor, supplier, or partner that does business with the United States military may conduct any commercial activity with Anthropic.”

But during Tuesday’s hearing, Hamilton conceded that Hegseth has no legal authority to bar military contractors from working with Anthropic on projects completely unrelated to the DoW. When Lin asked why Hegseth would publish the blanket ban, Hamilton replied, “I don’t know.”

Lin further pressed Hamilton on why the Pentagon rejected less punitive measures to transition away from Anthropic’s tools, noting that the supply chain risk designation is an extreme power almost exclusively reserved for foreign adversaries, terrorist groups, and other declared hostile actors.

Michael Mongan, an attorney with WilmerHale representing Anthropic, noted that the government’s use of this extreme designation against a simply “stubborn” negotiating partner is completely unprecedented.

The Pentagon says it is on track to replace all Anthropic technology with alternative AI tools from Google, OpenAI, and xAI over the coming months. It has also implemented safeguards to prevent Anthropic from tampering with systems during the transition. Hamilton told the court he did not know if it is even possible for Anthropic to push updates to its AI models without explicit Pentagon permission; Anthropic itself says it cannot make any unapproved changes to its government-deployed tools.

Separately, a ruling in Anthropic’s second lawsuit, pending before the federal appeals court in Washington, D.C., is expected to be issued shortly without an additional oral hearing.