A federal appeals court just gave the Pentagon a blank check to blacklist any AI company that won't toe the line — and Anthropic just learned the hard way that saying no to the military means losing billions.
This isn't about national security. It's about the military-industrial complex deciding who gets to participate in the AI future and who gets crushed. When the government can destroy a company's business because it won't grant "unfettered access" to its technology, that's not procurement policy — that's economic conscription.
The D.C. Circuit ruled 2-1 on Friday that the Pentagon was within its rights to designate Anthropic as a supply chain risk after the company refused to remove guardrails preventing its Claude AI from being used for autonomous weapons or mass domestic surveillance. Judge Gregory Katsas, a Trump appointee writing for the majority, bought the Pentagon's argument that "overly constrained AI models shutting down unexpectedly" could cause "important military operations to fail." He was joined by Judge Neomi Rao, another Trump appointee. Judge Karen LeCraft Henderson, appointed by George H.W. Bush, dissented.
Translation: if you won't let us use your tech for whatever we want — drone strikes, domestic spying, no questions asked — you're a threat to the republic.
Follow the money. Anthropic had a $200 million Pentagon contract signed in July 2025, CNBC reported. When negotiations over deploying Claude on the military's GenAI.mil platform collapsed in September, the Pentagon didn't just walk away — it weaponized two separate laws to blacklist the company entirely. Defense Secretary Pete Hegseth designated Anthropic under two different statutes, forcing the company to litigate in two separate courts. That's not regulation; that's punishment. And it's working. Anthropic says the designation has cost it "billions of dollars in lost business" and damaged its reputation ahead of a planned IPO, according to the New York Post. Meanwhile, the AI companies willing to play ball with no conditions get the contracts.
Trump made the power play explicit on Truth Social: "The Trump Administration has stopped AI 'people' from doing bad, or potentially bad, 'things,' like Dario (Anthropic!), who is now pretending to be a 'perfect little angel.'" CEO Dario Amodei's sin? Calling for an industry-wide slowdown and refusing to give the military a blank check on his technology. CNBC noted Amodei wasn't even invited to the White House state dinner for Chinese President Xi Jinping — because in Washington, access is a reward for compliance.
Here's the rub: a San Francisco federal judge already struck down the Pentagon's parallel designation under a different law last month, finding the administration had "unlawfully retaliated" against Anthropic for its views on AI safety. Two courts, two laws, two opposite conclusions. The New York Post framed the D.C. ruling as a straightforward "victory" for Trump and Hegseth; CNBC buried the conflicting San Francisco ruling deep in its story. Neither outlet wanted to say the quiet part out loud — the government is forum-shopping to crush a company that set its own terms.
The real question isn't whether Anthropic's safety concerns are right or wrong. It's whether a private company has the right to say no to the government — and whether the government can destroy you for exercising that right. The founders had a word for a government that conscripts private property and punishes dissent. They called it tyranny.
The Pentagon now has a playbook: demand unfettered access, blacklist the holdouts, and let the courts sort it out later. By the time any ruling comes, the target is already bleeding. That's not due process. That's a shakedown.
