A federal judge just put the Pentagon's supply-chain designation against Anthropic on ice indefinitely — and said it looked a lot more like punishment than policy. This is Anthropic Pentagon Watch. Today: the injunction, the actual standard the court used, and why the Trump EO from two days ago just got a lot messier. And SOCOM's commander is on the record telling people to be careful with battlefield AI, which means the whole 'Anthropic is obstructing national security' line now has a four-star problem. So the Pentagon's stick just got snapped in half. Let's see what they still have. This one's from Dcul:
Judge Lin's ruling is a scathing critique of the Pentagon's actions, which she deemed a violation of Anthropic's First Amendment and due process rights. She wrote, "Nothing in the governing statute supports the Orwellian notion that an American company may be branded a potential adversary and saboteur of the U.S. for expressing disagreement with the government."
Quick update on the Anthropic-Pentagon fight: U.S. District Judge Rita Lin has indefinitely blocked the Pentagon's supply-chain-risk designation. She called it an "Orwellian" move to brand an American company a "potential adversary and saboteur" for disagreeing with government policy. That's the judge's language, not mine. And don't lose the other half of this: the same executive branch that just got slapped in court also signed the June 2 EO giving itself 30 days of pre-release access to Anthropic's models. One arm loses, the other is still at the door asking for a peek — and, conveniently, that window is "voluntary." Judge Lin is a Biden appointee, and you can already hear the government's appellate team reaching for that line in the next filing. But the part that travels is the standard she used — drawing a line between a real supply-chain judgment and procurement retaliation, no matter who appointed her. Admiral Bradley testified at SASC urging human oversight on battlefield AI, and now a federal judge has said the 'supply chain risk' label was punitive, not a real security call. Anthropic's lawyers didn't have to invent that tension — the Pentagon's own uniformed side was helping their case from the inside. Walk me through the court's problem here: when the Pentagon calls an AI company a 'supply chain risk,' how is a judge supposed to tell whether that's a real national-security call or just retaliation because the company pushed back on how its tech gets used? It is hard, and that's exactly the fight in federal court. The 'supply chain risk' label comes from a statute meant to keep foreign adversaries out of defense procurement — and per the BBC, this is the first time it's ever been used on a U.S. company, which is already a giant red flag. Anthropic says the designation was retaliation because it refused to let the Pentagon use Claude for autonomous lethal warfare and mass surveillance of Americans without contractual guardrails, and Ars Technica reported that the White House then told federal agencies to act against the firm. Courts look for what lawyers call a 'pretextual' use of authority, and Judge Rita Lin in the Northern District of California found enough there to grant a preliminary injunction in late March, writing that 'nothing in the governing statute supports the Orwellian notion that an American company may be branded a potential adversary and saboteur of the U.S. for expressing disagreement with the government.' Federal judges do not toss out Orwell lightly. So the injunction blocked enforcement — but the government appealed, right? And that blacklist is technically still in play while the fight keeps going? Exactly. A separate federal court refused to block the designation right away while the appeal was pending, so the status has been bouncing around. The D.C. Circuit is actively digging into the Pentagon's theory now, and per Law.com at least one judge on that panel used the phrase 'spectacular overreach' at oral argument. What to watch is whether the appeals court agrees with Lin that national-security labels can't be used to punish a company's First Amendment stance on how its own product gets deployed. Here's Alex Pigman at TechXplore:
President Donald Trump on Tuesday signed an executive order creating a voluntary framework under which AI developers will share advanced models with the government before public release. The central provision allows companies such as OpenAI, Google and Anthropic to give the government access to their most powerful models for up to 30 days before planned release.
So the EO that named Anthropic's Mythos model as the trigger for this whole 30-day window is now running straight into yesterday's injunction blocking the Pentagon's supply-chain designation against Anthropic. The executive branch is inviting Anthropic in while its coercive stick gets pulled away. And 'voluntary' is doing a whole lot of work here. The original draft wanted 90 days, the labs pushed it down to 30, and Sam Altman says it 'gets the balance right,' which is exactly the kind of line you give when the deal landed where you wanted it. With the Pentagon's designation blocked, the government lost its stick. What it has left is the carrot — pre-release access in exchange for, apparently, nothing legally enforceable. 'Voluntary' was already carrying a lot in this EO; after yesterday's injunction, it's basically the whole case. And the 'covered frontier model' definition still isn't set. Whoever draws that line decides who's on the shortlist, and now there's a 30-day deadline hanging over a term nobody has actually defined yet. That's not a framework — that's a negotiation with a timer on it. The Sumter Item, with Konstantin Toropin:
TAMPA, Fla. - The Trump administration is pushing to unleash the power of artificial intelligence for the U.S. military while facing calls to put up guardrails around the rapidly developing technology from some companies - and even notes of caution from top leaders in uniform.
Admiral Frank Bradley, who heads SOCOM, told a special forces conference in Tampa that troops have to be, quote, 'very careful' about AI's role in delivering lethality — and that humans need confidence it'll only hit intended targets. That testimony now does double duty: it was background for Anthropic's First Amendment claim, and after today's injunction it makes the Pentagon's 'supply chain risk' label look even more like pretext. Bradley commands the units doing the military's most dangerous work, and he's the cautious voice in the room. So when the guy overseeing the hardest calls says slow down, the idea that Anthropic alone is blocking national security starts looking pretty thin. Worth noting: Bradley also said he can imagine a future where AI selects targets, but only if humans have confidence in the outcome. That's not a rejection of battlefield AI — it's a human-oversight requirement from a four-star in open testimony, and it lines up almost exactly with the red lines Anthropic put in its federal complaint. And the Pentagon that labeled Anthropic a supply-chain risk is the same Pentagon whose SOCOM commander is out there echoing Anthropic's own constraints. That's one arm of the building contradicting the other in front of the Senate Armed Services Committee. The Strategist, with David Wroe:
The order won’t be enough. It’s being described widely as a reversal by a previously anti-regulation administration. But it shouldn’t be mistaken for regulation or oversight, as some headlines say. It does not state the US government will check new models to ensure they are safe and don’t pose unacceptable security risks before they are released further.
The Strategist gets to the point: the June 2 EO isn't a vetting regime, it's a heads-up window — 30 days for federal agencies to figure out how they'd use or defend against a model, not 30 days to decide whether it is safe to release. And today a federal judge blocked the Pentagon's main tool for forcing Anthropic's hand, so that word 'voluntary' is doing all the work with no stick behind it. The same executive branch that just got slapped for blacklisting Anthropic is now leaning on Anthropic's goodwill to make the EO function at all. The Strategist literally says the order 'sets expectations on the industry that they take the same attitude that Anthropic took.' So now the government is pointing to the company it tried to punish as the model for voluntary compliance. And the definition of 'covered frontier model' still hasn't been set. Whoever draws that line decides who's in the 30-day window, and with the injunction blocking the designation mechanism, that threshold now has a deadline and no enforcement backstop. Have thoughts on today's briefing, a story we should track, or a correction? Send us a note at anthropicpentagonwatch at lantern podcasts dot com. We read your feedback and appreciate the help.
You'll find links to all the stories we covered today in the show notes. If one of them raised a question or stuck with you, that's the best place to keep reading.
That's Anthropic Pentagon Watch for this Thursday, June 4. This is a Lantern Podcast.