The Pentagon’s CTO just said the quiet part out loud — Anthropic is out, and that door is shut. Welcome to Anthropic Pentagon Watch. Today we’re pulling apart how fast the DoD is rerouting its AI money, what the classification fight actually exposes legally, and who’s slipping into the gap Anthropic left. And we’re naming the vendors getting paid while everyone else waves around safety theater, because “responsible AI” means nothing if the contract just lands with somebody else and a cleaner suit. It’s Friday. The receipts are in. Let’s read them. Here's Miranda Nazzaro at The Hill:
When Sanger brought up whether Anthropic could be part of these deals one day, Michael said, “not at the Department of War.”
Just more than two months have passed since the Pentagon blacklisted Anthropic from its military work after a dispute with the company over the potential use of its AI models for domestic surveillance or fully autonomous attacks.
Quick update on yesterday’s blacklist thread: the Pentagon’s CTO is now saying reconciliation is off the table. Emil Michael, the undersecretary for research and engineering, put it bluntly — “not at the Department of War.” That’s not a negotiation, that’s a lock clicking shut. And eight other companies — Microsoft, Google, Elon’s xAI — just signed classified-network deals last week. So no, the market didn’t sit around waiting for Anthropic. The Pentagon moved on and paid somebody else. Michael’s “single-threaded” line is the interesting procurement piece here. The Pentagon is turning vendor diversification into policy, which sounds fine until you realize it also means no single company can push back on a use case and expect to hold the line. Right — if you draw the line at autonomous weapons or domestic surveillance, the punishment is getting swapped out for somebody who won’t. That’s not procurement reform. That’s a compliance funnel. This one's from Noah News:
The US Department of Defense has told a federal appeals court that Anthropic’s fast-moving model development, combined with what it described as a breakdown of trust between the company and the Pentagon, supported its decision to classify the AI developer as a supply chain risk.
The Pentagon has told a federal appeals court that Anthropic’s rapid model iteration — plus what it’s calling a “breakdown of trust” — justifies classifying the company as a supply chain risk. That’s usually the kind of designation you see for foreign hardware vendors, not a San Francisco AI lab. And buried in the timeline is the part that matters: this seems to have started because Anthropic wouldn’t let its models be used for mass surveillance or fully autonomous weapons. So the Pentagon’s version is basically, you said no to us, now you’re the security threat. That’s not a trust breakdown, that’s a shakedown. Legally, this is genuinely uncharted. Supply chain risk designations carry real procurement consequences — blacklisting, contract exclusion — and they’ve never been stress-tested against a domestic AI company that just disagreed with the customer’s use case. I’ll give Anthropic this: holding a hard line on autonomous weapons and mass surveillance is exactly what their safety framing is supposed to mean in practice. The real question is whether they keep it when the legal pressure ramps up. Here's Insurance Journal:
The White House is looking into an executive order that would create a vetting system for new artificial intelligence models like Anthropic PBC’s Mythos in a bid to protect business and government networks from AI-related cyber risks, a top economic adviser said Wednesday.
The White House is floating an executive order to vet new AI models before release — basically FDA-style approval, but for systems that can find holes in your network. The model that kicked this off is Anthropic’s Mythos, which the company itself flagged as a potential global cybersecurity risk. Let’s be precise about what’s happening here: Anthropic built something dangerous enough to scare itself, quietly gave access to a few big banks and tech firms, and now the administration is running an “all of government” push to get it into federal systems. That’s not a safety protocol. That’s a procurement pipeline with extra steps. The FDA comparison from Hassett is doing a lot of work. FDA approval takes years and independent clinical trials. What he’s actually describing is a vetting roadmap he floated on Fox Business — those are not the same thing, and no order has been signed yet. And Anthropic gets to be the case study that shapes the whole vetting framework while its competitors are outside the fence. If you wanted to build a regulatory moat, this is exactly how you’d do it. TRT World writes:
The United States military has taken a defining step towards becoming what it calls an " AI-first fighting force." This month, the Pentagon announced agreements with eight of America's most powerful technology companies to deploy artificial intelligence directly onto its most classified military networks.
Eight AI firms just got keys to the Pentagon’s most classified networks — Impact Level 6 and 7, which is about as deep as it gets in DoD infrastructure. Google, Microsoft, OpenAI, Amazon, the usual names, plus a startup called Reflection that most people haven’t heard of. A startup called Reflection on a classified military network — that’s not a watershed moment, that’s a procurement officer’s headache waiting to happen. And notice Anthropic is not on that list. Anthropic drew a public line against autonomous weapons and mass surveillance use cases, which is either a real ethical constraint baked into the contract, or a very effective way to stay out of the legal blast radius when something blows up downrange. I’ll believe the ethical line when I see an enforcement mechanism. Right now it’s a press release. Who decides when a targeting system crosses into “autonomous” — the company, the contracting officer, somebody else? Nobody wants to answer that. From Tastytech:
The phrase “any lawful use” formed the centre of the recent disagreement between Anthropic AI and the US administration, with CEO Darius Amodei claiming that it would let the US government use Anthropic technology to subject the American civilian population to surveillance, and produce autonomous weapons, areas of Anthropic’s use that he wanted walled off.
The Pentagon just added four more AI suppliers to its classified-ops roster — Microsoft, Amazon, Nvidia, and Reflection AI, a company that hasn’t even released a public model yet. That brings the approved list to seven, and Anthropic is still not on it. Right, because Anthropic pushed back on “any lawful use” — specifically domestic surveillance and autonomous weapons — and the Pentagon answered by canceling a two-hundred-million-dollar contract and tagging them as a supply chain risk. That’s the first time that label has been applied to a U.S. company. It’s not a legal finding, it’s a message. Anthropic sued, saying it lost money not just on the canceled contract but from downstream clients who got spooked by the “supply chain risk” label. Whether the courts treat that as actionable government interference or just aggressive procurement politics — that’s the live question. And now the Pentagon has seven vendors willing to sign off on “any lawful use.” So the leverage move worked. The message to every other lab is simple: put the limits in writing, and we’ll replace you with someone who won’t. If you’re tracking AI power and government accountability here, try Musk v Altman Daily — daily court-watch on Elon Musk’s trial against Sam Altman, OpenAI, and Microsoft, including testimony, exhibits, and the AGI governance fight. Find it wherever you listen to podcasts.
If you want to dig deeper, we’ve put links to all of today’s stories in the show notes. Follow up on the ones that grabbed you.
That’s Anthropic Pentagon Watch for Friday, May 8th. Thanks for listening, and have a good weekend. This is a Lantern Podcast.