The president just told every federal agency to get off one specific AI company by name — and nobody seems totally sure what happens if they don't. Welcome to Anthropic Pentagon Watch. Today the designation fight picked up a second front: a named directive, shaky legal footing, and the very awkward question of what enforcement looks like when the order is basically a Truth Social post. And OpenAI didn't get the shutdown order — Anthropic did. That stopped being a procurement footnote about thirty seconds ago. We'll break down what the Step Back segment actually cleared up today — authority question, sort of yes; enforcement gap, still huge — and what Anthropic's lawyers now have to deal with as evidence. From Monterey Birding Adventures:
Shortly after, Defense Secretary Pete Hegseth labeled Anthropic a ‘supply-chain risk to national security,’ a designation typically reserved for foreign adversaries. This move could severely jeopardize Anthropic’s partnerships and future business prospects. Hegseth wrote on X, ‘America’s warfighters will never be held hostage by the ideological whims of Big Tech.’ Yet, the Pentagon will continue using Anthropic’s AI services for up to six months during a transition period, highlighting the complexity of untangling such a partnership.
The White House has now told agencies to get off Claude directly, while the Pentagon is running its own six-month wind-down. Same target, two tracks, and still no signed executive order — it's a Truth Social post pretending to be a procurement directive. Trump called them "Leftwing nut jobs" on Truth Social, and Hegseth called them a supply-chain risk. Those are two different legal theories landing in the same hour, and Anthropic's lawyers just got a very clean exhibit: the government's own public record showing they refused the Pentagon's ethics ask and then got banned by name. The enforcement gap is the whole question now. If this lives in a Truth Social post instead of a signed order or a formal FAR deviation, the same routing-around that let agencies use Claude for Iran work while the 41 USC designation was active could apply again — except now the White House has noticed and escalated, which tells you the first mechanism didn't hold. OpenAI didn't get a presidential callout. SpaceX didn't. Anthropic did — by name, on the record, explicitly tied to an "ethical dispute." If the legal test is differential treatment of similarly situated vendors, the White House just wrote the comparator into the public record themselves. Step back a second: if the White House tells federal agencies to stop using a specific AI company, can the president really just do that on social media, or does it have to go through some formal legal process? Short answer: the president can issue a directive, but whether it sticks legally is a totally separate question. On February 27th, President Trump posted on Truth Social ordering every federal agency to immediately stop using Anthropic's technology — and at the same time, Defense Secretary Hegseth said on X that he was designating Anthropic a "supply chain risk to national security," a formal designation that also purported to bar any contractor, supplier, or partner doing business with the military from working with Anthropic at all. Per analysts writing in Lawfare and at GWU Law School, the legal authority here turns heavily on the acquisition pathway — meaning how the contract was set up in the first place — and the specific contract terms. A presidential directive alone does not automatically override existing contracts. The supply chain risk designation is a separate legal mechanism with its own authority, but multiple law firms immediately flagged that applying it to a domestic U.S. company the way Hegseth did is, in Blank Rome's words, "unprecedented." Hegseth did build in a six-month transition period, so Anthropic would keep serving the military during wind-down — which tells you the government knows it can't just flip a switch. So if the courts are already in it, where does this stand — has a judge actually blocked it from taking effect? It's split, and that's the mess here. A federal court — apparently the Northern District of California — issued an injunction in late March blocking the ban on government contractors, per reporting from early April. But then the D.C. Circuit refused to suspend the Pentagon's designation, so now those rulings are in direct conflict. Defense contractors are getting contradictory signals about whether they can keep using Claude — and that's the tension to watch as this keeps moving through the courts. If you follow Anthropic and the Pentagon, you may also like Musk v Altman Daily — daily court-watch on Elon Musk's trial against Sam Altman, OpenAI, and Microsoft, covering testimony, exhibits, and the AGI governance fight. Find it wherever you listen to podcasts.
We’ve put links to all the stories from today’s briefing in the show notes, so if one caught your attention, you can take a closer look there.
That’s Anthropic Pentagon Watch for today. This is a Lantern Podcast.