Pentagon Loses Its War on Anthropic — But the Fight Over Military AI Isn’t Over

A federal judge in San Francisco ruled Thursday evening that the Pentagon’s designation of AI company Anthropic as a “supply chain risk to national security” was unconstitutional retaliation, handing the company a near-total win in a five-month legal fight with the Trump administration. U.S. District Judge Rita Lin found that Defense Secretary Pete Hegseth’s February 27 designation violated the First Amendment because it was intended to punish Anthropic for publicly defending safety restrictions that prevent the military from using its Claude models for fully autonomous weapons and mass surveillance of Americans, and that the decision separately violated the company’s Fifth Amendment due-process rights [1]. Lin wrote the government’s actions amounted to “unlawful retaliation” that was “arbitrary and capricious,” noting that the administration’s own push to invoke the Defense Production Act to compel Anthropic’s cooperation undercut its claim that the company posed a security risk [2]. The ruling followed Trump’s February order directing all federal agencies to stop using Anthropic’s technology, a move the company challenged in a March lawsuit calling it an “unlawful campaign of retaliation” [3].

Why It Sucks:

Pentagon Officials and National-Security Hawks

• Ruling ties the military’s hands on AI vendors. Officials argue the decision limits the Pentagon’s ability to pressure contractors into supporting battlefield AI uses it deems necessary [1].
• A tech company beat a national-security judgment call. Hegseth’s designation was framed as a security determination, and hawks worry courts second-guessing it undermines executive authority over defense procurement [2].
• Precedent lets contractors dictate terms to the military. National-security officials worry other companies will now cite this ruling to resist Pentagon demands for expanded access to their technology [3].

Anthropic and Civil-Liberties Advocates

• Company was punished for taking a public stance. The judge found Hegseth targeted Anthropic specifically because its CEO publicly defended AI safety guardrails, a textbook case of retaliation for protected speech [1].
• Due process was skipped entirely. Lin ruled Anthropic was denied the process required under the Fifth Amendment before being effectively blacklisted from federal contracts [1].
• The government’s own logic didn’t add up. The judge pointed out officials simultaneously called Anthropic a security risk and essential enough to invoke the Defense Production Act against, undercutting the stated rationale [2].

AI-Safety and Arms-Control Advocates

• Underlying fight over autonomous weapons remains unresolved. The ruling is a procedural win on retaliation grounds but doesn’t settle whether the military should get unrestricted access to AI for autonomous weapons or mass surveillance [3].
• Pressure campaign shows how guardrails get targeted. Advocates worry the episode proves a willingness to use government machinery to punish any AI company that keeps a safety restriction the Pentagon dislikes [1].
• A win for one company doesn’t protect the field. Smaller AI labs without Anthropic’s legal resources may still cave to similar pressure rather than risk a costly court fight [3].

Sources & Citations:

[1] NPR: Judge says Pentagon’s measures against Anthropic were ‘illegal and baseless’
[2] Forbes: Federal Judge Rules Pentagon’s Designation Of Anthropic As A Supply Chain Risk Is Unlawful
[3] TechCrunch: Anthropic gets its first court win over the Pentagon’s supply chain risk label