The Pentagon spent the week fighting in court to keep calling Anthropic a supply-chain risk — and signing classified deals with everyone but Anthropic. If you're just joining, this fight started over Claude's use restrictions. Anthropic says it won't support lethal autonomous warfare or mass domestic surveillance. The Pentagon responded by ending contracts, slapping on a supply-chain-risk label with government-wide consequences, and demanding broader access. A district judge ordered agencies not to enforce that label, while separate national-security controls have already forced Anthropic to pull its newest Fable and Mythos models offline worldwide. You're on Anthropic Pentagon Watch. Today this stops being a negotiation — three of Anthropic's biggest cloud rivals just walked off with the classified contracts. We'll name who got paid. If Anthropic-Pentagon supply-chain-risk fight matters to you, hit follow — we'll be back on it soon. Here's Theresa Maher at Inside Defense:
The Pentagon told the California Northern District Court that it will be bringing U.S. District Court Judge Rita Lin’s March 26 ruling -- to prohibit federal agencies from designating Anthropic a “supply chain risk” and cutting the company’s technology from all federal systems -- to the Ninth Circuit Court of Appeals for reconsideration.
Here's the update on the supply-chain-risk fight: the Pentagon is taking Judge Rita Lin's March 26 injunction to the Ninth Circuit. Lin told federal agencies they couldn't brand Anthropic a 'supply chain risk' and then rip its tech out of every federal system. Now the Pentagon has filed a notice of appeal to keep that label alive. Notice the timing — there was already a seven-day stay on Lin's order. Under Secretary Emil Michael basically said the stay existed to give them runway for an appeal. So nobody's surprised here. And here's the part to say plainly — they're trying to preserve the precedent in court while they sign replacement contracts elsewhere. Two tracks: keep the label alive legally, swap the vendor in practice. Right. The dollars already moved. The appeal is a doctrine play, not a push to get Claude back into systems. They want a court to bless the idea that the military can call an AI lab a national-security threat and make it stick. The Washington Post, with Tara Copp and Ian Duncan:
Anthropic said late Thursday that it will not concede to the Pentagon’s ultimatum for full access to its artificial intelligence tool Claude, saying it cannot loosen its restrictions against use in fully autonomous weapons or mass domestic surveillance. Anthropic’s announcement comes hours before a Pentagon-imposed Friday deadline to comply — setting the stage for the AI firm to either win a last-minute concession by Defense Secretary Pete Hegseth or face the possibility of being cut from future military work.
Dario finally said it — out loud, on the record: domestic mass surveillance and fully autonomous weapons. Those are the two lines Anthropic won't cross. And the Pentagon's reply, per the Post, is that it never considered either use in scope. Which raises the obvious question — if those uses were never on the table, why not put the ban in the contract? Right? You can't say 'we'd never do that' and then say, 'but we refuse to promise we won't.' The second sentence tells you more than the first. The mechanics here matter. Amodei went public hours before Hegseth's Friday deadline — a refusal with a clock on it, not a negotiation. And press them on what 'won't permit' even means. Is that a policy line in a terms-of-service doc, or an actual technical hard stop in the model? Because the Pentagon's shrug suggests they think it's neither. The Cato angle gets sharper here too — Amodei spent months arguing governments should be able to block dangerous AI. Today he's the one doing the blocking, and the government's getting told no. Here's Ian Duncan at The Washington Post:
Eight leading artificial intelligence companies have reached deals to deploy their technology in classified Pentagon computer networks, the Defense Department said Friday. The announcement leaves Anthropic — one of the industry’s key players — increasingly isolated as it battles the Trump administration in court over being branded a national security risk by the Pentagon.
Eight companies signed onto classified networks Friday — Amazon, Google, and Microsoft among them. In the Defense Department's telling, Anthropic is the holdout; in the Post's framing, it's increasingly isolated. There's your who-gets-paid answer. Anthropic says no on autonomous weapons and domestic surveillance, and three of its biggest cloud rivals walk into the classified room the same week. And put that next to the supply-chain appeal we just covered — the Pentagon's fighting to keep the risk label alive while it signs replacements. The precedent stays in court; the vendor changes in practice. The message every other lab hears: hold a guardrail, lose the contract. That's the price tag, spelled out in eight signatures. The hard question for Anthropic — did Amodei's whole regulatory push cost him the access that would've let him actually shape the rules? He wanted binding federal AI standards. The government answered through procurement instead. And here's what I want someone pressing them on — are Amazon, Google, and Microsoft offering the same safety constraints Anthropic refused to drop, or are they selling access without them? Eight companies are in classified networks, and nobody's said whose models stop at the lethal line. Thomas A. Berry, Addison Bennett, Sopen Shah, and Sarah Grant, writing in Cato Institute:
And this stalemate came to a head in February of this year, when the Department not only ended its contracts with Anthropic (subject to a six-month wind-down period) but also labeled Anthropic a “Supply-Chain Risk to National Security.” The latter designation forbids any business that contracts with the federal government from working with Anthropic in any capacity, devastating Anthropic’s business.
So here's where the Cato brief stands today: filed March 9, arguing Anthropic's model design choices — including the no-autonomous-weapons, no-mass-surveillance lines — are First Amendment-protected expression. That's the constitutional theory. And right now, it's the only lane Anthropic has left that isn't already behind on points. Right, because everything else cratered this week. The appeal we hit earlier, the classified deals going to Amazon, Google, and Microsoft — Cato's brief is the one card left on the table. And it runs straight into the Pentagon's posture. DoD is in court arguing it has the right to keep calling Anthropic a supply-chain risk. Cato says the thing they're being punished for is protected speech. Both arguments can't win. And I love the framing here, honestly — Anthropic's refusal as expression. The product decision is the message. Whether a panel that already denied the stay buys that? Different question. So that's the test now — does this brief change the appellate math, or just add one more theory for a court that's already leaning the other way? If you're following the Pentagon's AI moves, you might also like The Data Center Daily, a daily briefing on AI compute, hyperscaler capex, the power grid, semiconductor supply, and energy markets. Find it wherever you listen to podcasts.
Next, we're watching the Ninth Circuit review of Judge Rita Lin's March 26 order blocking the supply-chain-risk designation.
If you want to dig further into anything we covered today, you'll find links to every story in the show notes. Follow whichever threads caught your ear.
That’s Anthropic Pentagon Watch for today. This is a Lantern Podcast.