← Anthropic Pentagon Watch

Anthropic’s Pentagon Fight Turns on Procurement and Speech (June 22, 2026)

June 22, 2026 · 6m 57s · Listen

The Pentagon labeled Anthropic a supply-chain risk. Now Cato says that label runs into the First Amendment — and that's a much tougher wall to climb. This is Anthropic Pentagon Watch. Today: a brand-new legal theory enters the fight, and somehow, the same dispute is alive in two courthouses at once. We'll start with what Cato's actually arguing. Tap follow so the next episode finds you. Here's David Mortlock; Britt Mosman; David Levine at Willkie Farr & Gallagher LLP:

Designation of an American company as a supply chain risk under 10 U.S.C. § 3252 appears unprecedented—the provision has previously been used to target foreign adversaries such as Russia’s Kaspersky Labs and China’s Huawei. Use of Section 3252 typically requires official findings by contracting and information security officers; it is unclear whether the Defense Department followed any such process here.

Look at the date on this Willkie Farr memo — March 2. That's five days after Trump's Truth Social post, and well before Judge Lin's injunction ever existed. So a third major firm was billing contractor clients to tell them how to rip Claude out, before a single judge had weighed in. The compliance machine didn't wait for a court. And notice what the memo quietly admits — neither Trump's directive nor Hegseth's tweet cited any legal authority. Anthropic had to go find the statute for them. The statute is 10 U.S.C. § 3252 — it lets DoD exclude designated systems from procurement. So yes, the authority exists. It just wasn't invoked by anyone actually issuing the order. Which means the order that triggered the entire scramble was a tweet from SecWar. And firms were charging clients to comply with a tweet. Step back for me: if Anthropic and the Pentagon are now fighting in different courts, how is the same dispute alive in multiple places at once — and how do we tell whether a ruling is a real win on the merits, or just procedural shadowboxing? Right. The short answer: they're two legally distinct cases, even though they come out of the same fight. Anthropic put contract limits on Claude — no surveillance of Americans, no powering fully autonomous weapons. The Trump administration said it needed Claude for, quote, 'all lawful purposes.' When they couldn't settle that, the Pentagon designated Anthropic a supply-chain risk in early March, basically branding it a national-security threat, per CNBC. That designation kicked off the parallel litigation. In the Northern District of California, Judge Rita Lin issued a preliminary injunction in late March blocking enforcement of the administration's directive to stop using Claude. She wrote that the government was trying to 'cripple Anthropic' and 'chill public debate,' and Lawfare noted she found the actions retaliatory for First Amendment-protected expression. Separately, the challenge to the supply-chain designation went to the D.C. Circuit. In April, that court denied Anthropic's request for a stay, so the designation stays on the books for now, as WIRED reported. The key word in both outcomes is 'preliminary.' Neither court has decided who's right on the merits. Preliminary injunctions and stay denials are about whether to hit pause while the real case plays out. They're high bars, but they're not final verdicts. So when the D.C. Circuit denied the stay, that didn't actually contradict what the San Francisco judge did — they were looking at different things? Exactly. Jones Walker's legal analysis put it cleanly: the two courts were in 'two different postures' on two different legal questions. So the preliminary rulings can point in different directions without canceling each other out — though WIRED noted it wasn't immediately clear how they'd ultimately be reconciled. Next up, a D.C. Circuit panel heard oral arguments in May, and per AP, the judges sounded divided. A substantive appellate ruling on the supply-chain designation could be the first real merits decision in this whole saga. From Thomas A. Berry, Addison Bennett, Sopen Shah, Sarah Grant at Cato Institute:

But Anthropic consistently drew two red lines: it would not provide AI services to aid in lethal autonomous warfare or mass surveillance of Americans. The Department grew increasingly frustrated with Anthropic’s refusal to depart from these policies. 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.”

Okay, here's the escalation. Every brief so far has fought this on contract law and due process — basically, did the Pentagon follow its own procurement rules. Cato just walked in with a completely different theory: Claude's guardrails are speech. Berry and three co-counsel make that argument in a March 9 filing: Anthropic's design choices — refusing lethal autonomous warfare, refusing mass surveillance of Americans — are First Amendment-protected expression. And nobody's said the obvious next sentence. If the guardrails are speech, then the Pentagon ordering Anthropic to strip them out is compelled speech. That's a much harder wall for the government to climb than 'we didn't give you enough notice.' The district court already called the supply-chain label Orwellian on due process grounds. Now Anthropic has two runways — procedural and constitutional. We'll see whether they reinforce each other on appeal or pull in opposite directions. Right — if the Ninth Circuit is cold on the due process path, does the First Amendment theory give Anthropic a second shot, or does running both make the posture messy? Genuinely don't know yet, and that's why this brief matters. Legally, treating model behavior as protected expression is a stretch with very little precedent. Cato knows that. They're planting a flag for the merits while everyone else litigates paperwork. Got feedback, a story idea, or a correction for us? Send a note to anthropicpentagonwatch at lantern podcasts dot com. We read the inbox, and your tips help shape what we cover next.

We’ve put links to every story from today’s briefing in the show notes. If something stuck with you, you can dig into the source material there. Thanks for spending part of your Monday with us. That’s Anthropic Pentagon Watch for today. This is a Lantern Podcast.