A Federal Judge Ruled the Pentagon's Anthropic Blacklist Was Illegal

A federal judge ruled that the Pentagon unlawfully labeled Anthropic a supply chain risk after the AI company refused to allow its Claude model to be used for mass domestic surveillance or fully autonomous weapons. U.S. District Judge Rita F. Lin's 59-page decision, issued Thursday, found the designation was retaliation for Anthropic's public criticism, not a genuine security threat, and ordered the government to reverse related directives. The ruling stems from a dispute over a $200 million Department of Defense contract that Anthropic walked away from in February.

A federal judge has ruled that the Pentagon broke the law when it branded Anthropic a supply chain risk. The ruling doesn't force Washington to use Claude, but it does say the government can't turn a contract fight into punishment for public criticism. A federal judge in San Francisco has handed Anthropic a serious win against the Pentagon. U.S. District Judge Rita F. Lin ruled Thursday that the Trump administration acted unlawfully when it labeled the AI company a supply chain risk after a fight over how Claude could be used by the military. According to AFP, Lin wrote in her 59-page decision that "the empty invocation of national security is not a blank check to punish and retaliate against government critics." That's the case in one sentence. The ruling matters because this was never only about one vendor contract. If you sell software into government, especially AI software, the Anthropic fight is a warning about how quickly procurement pressure can turn into political pressure. Lin's order says there is still a line there. The blacklist followed a fight over Claude The Pentagon's supply chain risk label is usually aimed at companies Washington believes could compromise government systems. Anthropic is not Huawei. It's a San Francisco AI company that had already been working with the federal government, including through a two-year Department of Defense agreement with a $200 million ceiling announced by Anthropic in July 2025. Google DeepMind Is Losing the AI Talent War It Used to Win Easily https://startupfortune.com/google-deepmind-is-losing-the-ai-talent-war-it-used-to-win-easily/ Google DeepMind's arrivals-to-departures ratio has fallen from about 12-to-1 in 2023 to roughly 2-to-1 in the third quarter of 2026, according to Fortune. Jeff Dean, Sanjay Ghemawat, Noam Shazeer, and Nobel laureate John Jumper are among the recent departures to rivals and startups, with Anthropic pulling in the largest share. - why top AI researchers are leaving Google DeepMind https://startupfortune.com/google-deepmind-is-losing-the-ai-talent-war-it-used-to-win-easily/ - Google DeepMind losing talented scientists to competitor companies https://startupfortune.com/google-deepmind-is-losing-the-ai-talent-war-it-used-to-win-easily/ The fight started over two uses Anthropic would not allow. In a February statement, the company said it supported lawful national security uses of Claude but would not permit the model to be used for mass domestic surveillance of Americans or fully autonomous weapons. Those were the hard lines. The Pentagon wanted more room. Axios reported in February that Defense Secretary Pete Hegseth gave Anthropic CEO Dario Amodei a Friday deadline to give the military broader access to the company's technology or face penalties, including a possible supply chain risk designation. The Washington Post reported the same week that Anthropic was prepared to walk away from the $200 million contract if those concerns were not addressed. It did walk away. Then the fight moved to court. Lin had already signaled where she stood in March, when she granted Anthropic a preliminary injunction. In that earlier order, she wrote that nothing in the statute supported the "Orwellian notion" that an American company could be branded a potential adversary and saboteur for disagreeing with the government. Thursday's ruling made the core finding permanent and ordered the government to reverse directives tied to the designation, according to the Wall Street Journal. The court saw retaliation, not risk Frankly, the Pentagon's problem was the paper trail. Lin found that the government's stated security rationale did not match the record in front of her. The court pointed to government material describing Anthropic as engaging in an "increasingly hostile manner through the press," language that tied the designation to the company's public criticism rather than to a technical threat. That's not a supply chain problem. That's a speech problem. The distinction matters for every AI lab now negotiating with Washington. The government can choose not to buy a model. It can demand contract terms. It can decide that a vendor's policies are too restrictive for a mission. What it can't do, Lin's ruling says, is use a national security label as a punishment because a company refused to drop its public position. 80% of Developers Say AI Coding Tools Feel Like Dependence, Not Help https://startupfortune.com/80-of-developers-say-ai-coding-tools-feel-like-dependence-not-help/ A new Coddy Developer Survey, reported by ZDNet, finds 80% of developers now describe their AI coding tool use as dependence rather than advantage. Forty-three percent keep coding after hours despite meaning to stop, yet 74% still believe heavier AI use will boost their odds of promotion. - developer dependence on AI coding tools survey https://startupfortune.com/80-of-developers-say-ai-coding-tools-feel-like-dependence-not-help/ - AI coding habits creating dependency among developers https://startupfortune.com/80-of-developers-say-ai-coding-tools-feel-like-dependence-not-help/ Anthropic still isn't out of the legal woods. The Information reported after the March injunction that a separate Pentagon designation under a different supply chain statute was being challenged in a Washington case, and the Wall Street Journal reported Thursday that another pending case could offer the company less support. So the California ruling is a major win, but not the final word on every government restriction Anthropic faces. Claude also didn't disappear while the Pentagon fight played out. TechCrunch reported in March, citing Sensor Tower data, that Claude rose from just outside the App Store's top 100 at the end of January to the No. 1 free app in the U.S. App Store after the dispute became public. That doesn't settle the legal question, but it does puncture the idea that Washington could quietly isolate the company and watch the market follow. OpenAI, Google, Meta and every smaller defense AI contractor should read this ruling closely. The government remains the buyer with the biggest checkbook and the deepest security claims. But a buyer is not a king. If Washington wants access to frontier models, it has to negotiate for it, not dress retaliation up as risk management. Also read: Google's AI Mode Now Tracks Flight Prices and Books Your Hotel https://startupfortune.com/googles-ai-mode-now-tracks-flight-prices-and-books-your-hotel/ • Researcher Alon Hertz Tricked Claude, Codex and Hermes Into Running Malware https://startupfortune.com/researcher-alon-hertz-tricked-claude-codex-and-hermes-into-running-malware/ • Trump Weighs Chip Tariffs Up To 100% That Could Slow The AI Buildout https://startupfortune.com/trump-weighs-chip-tariffs-up-to-100-that-could-slow-the-ai-buildout/