Anthropic beat the Pentagon’s AI blacklist—for now. And “for now” is where the billable hours live. If you’re just joining us: Anthropic has been fighting the Pentagon over supply-chain-risk authority, military-use restrictions, and contractor access to Claude. The dispute started when Anthropic refused certain military uses, including surveillance and autonomous weapons, while the Pentagon’s classified-systems AI agreements excluded the company. Judge Rita Lin had already questioned whether the blacklist was punishing Anthropic’s policy stance instead of addressing a real security risk. This is Anthropic Pentagon Watch. A court just drew a line, and Anthropic’s already hiring to get back into the market that tried to freeze it out. Here's Osmond Chia at BBC News:
A judge has ruled the US Department of Defense acted unlawfully when it designated artificial intelligence (AI) startup Anthropic a supply chain risk, calling the move "illegal and baseless". The company argued in the lawsuit that Defense Secretary Pete Hegseth overstepped his authority with the new designation.
Update on the Anthropic-Pentagon supply-chain-risk fight: Judge Rita Lin has ruled the blacklist was unlawful retaliation. “Illegal and baseless” is a considerably less flattering review than the Pentagon was hoping for. The Defense Department used a designation usually aimed at foreign threat companies against a U.S. lab because it balked at mass surveillance and autonomous weapons. Judge Lin saw it as punishment. Lin’s key point: national security is no blank check for retaliation against critics. That puts the contractor-boycott order in a very ugly light—and any replacement exclusionary policy needs an actual, independent rationale. And Anthropic is already hiring for national-security sales. Fine—sell your tools. Just don’t use a court win over limits on surveillance and killer robots to cover a quiet sprint back to defense money. So, does this ruling say Anthropic has a constitutional right to keep Claude out of autonomous weapons or mass surveillance? Or is it narrower: the Pentagon can’t punish the company for setting those limits? And could the government still pick another AI provider with a real procurement or security rationale? Mostly the second. The court found that the government unlawfully retaliated against Anthropic for protected speech; it did not declare that Anthropic can constitutionally control every military use of AI. CBS News reports that Judge Rita Lin permanently barred enforcement of the measures that would have cut Anthropic off from the federal government because the administration had unconstitutionally punished the company for protected speech. The dispute grew out of Anthropic’s effort to keep Claude from being used for mass surveillance or fully autonomous weapons, while the military argued it should be able to use the model for all lawful purposes. CNN reports that the judge ordered the Pentagon’s supply-chain-risk designation removed and called the action unlawful retaliation. The military still has ordinary discretion over whom it contracts with: Judge Lin also said it should have wide latitude in deciding which companies to work with. The constitutional line is whether a decision rests on procurement or a real security concern—or whether contracting power is being used to punish a company for criticizing the government or maintaining public safeguards. But “national security” is such a broad label. What stops the Pentagon from getting to the same result again with a more carefully written justification? The ruling doesn’t erase the Pentagon’s ability to make national-security or procurement judgments. It says that rationale can’t be a pretext for retaliation. As The Guardian reports, Judge Lin wrote that an “empty invocation of national security” is not a blank check to punish government critics. Now watch whether any future restriction is tied to a documented, independent risk—or again tracks Anthropic’s speech and its limits on military uses of Claude. Open Magazine writes:
The disagreement destroyed a $200 million alliance, provoked an ultimatum from US Defence Secretary Pete Hegseth and transformed Anthropic from an important military technology supplier into an alleged threat to America’s supply chain. President Donald Trump directed federal agencies to stop using the company’s technology. Hegseth announced that contractors working with the military would have to sever commercial relationships with Anthropic.
A $200 million Pentagon alliance blew up because Anthropic wouldn’t allow mass domestic surveillance or weapons to choose human targets without meaningful human control. Hegseth’s answer was: take every lawful use, or get out. Very subtle negotiation style. And Open Magazine makes the procurement fallout concrete: Trump ordered agencies off Anthropic, while contractors were told to sever commercial ties. Judge Lin’s merits ruling leaves that contractor-pressure campaign looking very exposed. The court stopped retaliation; it didn’t resurrect that $200 million deal or reserve Claude a seat in every classified program. The Pentagon can still buy from somebody else if it has an actual independent reason—instead of stapling “supply-chain risk” onto punishment. The ruling is substantial, but it’s still a district-court decision and can be appealed. So now the government’s lawyers get the unpleasant assignment: find a clean procurement rationale, if one exists, without recreating the retaliation in a nicer font. Zachary Groz, writing in The American Prospect:
This week, artificial intelligence giant Anthropic began advertising a new job to lead its “National Security Sales” team, according to a public posting reviewed by the Prospect. The job opening, which hasn’t been previously reported, signals that the end may be in sight for Anthropic’s extended spat with the Department of Defense and the Trump administration, which earlier this year blacklisted the company from doing business with the federal government or any federal contractors, forever.
The American Prospect found Anthropic recruiting a Head of National Security Sales—up to $700,000 a year—to rebuild business with the military and intelligence agencies. The court took retaliation off the table, and Anthropic is already looking for someone to get it back through the door. Very efficient timing. Anthropic defended its right to draw lines around autonomous weapons and mass surveillance, then posted a senior sales role for the very institutions testing those lines. Fine—call it a business decision, because it is. But “safe, frontier AI” is a sales slogan until Anthropic says what its new national-security team will refuse to sell for surveillance pipelines or weapons targeting. The ruling we just covered doesn’t award Anthropic a contract or a slot on classified systems. It bars punishment for protected speech; DoD can still build a genuinely independent procurement case. Expect a great many lawyers to discover the word “independent.” If your team needs this kind of briefing for your own industry, Lantern makes private daily podcast briefings on your competitors, market, or beat, delivered to a private feed for your whole team. Learn more at lantern podcasts dot com slash briefings, with a 14-day free trial.
Links to every story we covered are in the show notes if you’d like to dig into the details or read more about the ones that caught your attention. That’s Anthropic Pentagon Watch for today. This is a Lantern Podcast.