Judge to Pentagon: Calling AI 'Woke' for Refusing Killer Robots Isn't a National Security Strategy

A federal judge has just issued a reality check to the Pentagon, slapping a preliminary injunction on its attempt to brand AI startup Anthropic a "supply chain risk." U.S. District Judge Rita Lin (N.D. Cal.) ruled Thursday that the government's move was about as legally sound as a rug pull, violating Anthropic's First Amendment and due-process rights. She pointed out that the statute in question doesn't support labeling a domestic company an adversary just because it told the government "no."

The court dropped this judicial mic just two days after oral arguments, with the judge noting the government's own paperwork basically admitted to a retaliatory motive. Public-affairs attorney Andrew Rossow, CEO of AR Media Consulting, told Decrypt the designation was "triggered by press conduct, not a security analysis"—or, in degen terms, it was a pure emotional trade, not based on fundamentals.

This whole saga started with a two-year, $200 million contract from the Department of War's Chief Digital and Artificial Intelligence Office, awarded in July 2025. Talks to put Anthropic's Claude on the GenAI.Mil platform hit a wall when the company, citing safety concerns, insisted its model not be used for mass surveillance of Americans or lethal autonomous warfare—a stance more principled than your average governance token voter.

At a Feb. 24 meeting, Secretary of War Pete Hegseth gave Anthropic a three-day ultimatum to drop its ethical restrictions, threatening an immediate supply-chain-risk label for non-compliance. In a stunning coincidence, former President Trump posted on Truth Social the same day, demanding every federal agency "immediately cease" using Anthropic's tech and calling the firm "radical left" and "woke." Hegseth later called Anthropic's stance a "master class in arrogance and betrayal" and banned any military contractor from working with them, effectively trying to deplatform the AI firm from the defense industrial complex.

The formal "you're a risk" designation arrived via a March 3 letter. Anthropic, refusing to take that sitting down, sued on March 9, alleging violations of the First Amendment, due process, and the Administrative Procedure Act. Judge Lin's order, stayed for seven days, blocks all three government actions, demands a compliance report by April 6, and restores the pre-Feb. 27 status quo—a full revert to a previous, less chaotic state.

Historically, "supply chain risk" tags were reserved for foreign intelligence agencies, terrorists, and other hostile actors—never a domestic company. It's like using a ban hammer meant for scammers on a project that just refused a VC's term sheet. Unsurprisingly, defense contractors began reassessing, and in many cases terminating, their ties with Anthropic after the label dropped, showing that FUD works even in the world of billion-dollar contracts.

Policy strategist Pichapen Prateepavanich of Gather Beyond told Decrypt the ruling could encourage AI firms to actually codify their ethical guardrails when dealing with governments. It shows companies can set clear usage limits without immediately getting hit with punitive regulatory action—though the underlying tension between building in public and dealing with the state remains, a classic builder vs. regulator showdown.

Rossow warned that the government's legal theory is a straight-up "weaponization" of the law, part of a pattern of executive overreach resulting in "disproportionate, emotionally-driven and biased threats." He cautioned that accepting this theory would set a dangerous precedent, allowing AI firms to be blacklisted for safety policies the government dislikes, all under the convenient, all-purpose banner of "national security." It's the regulatory equivalent of calling something a security to justify any action you want.