The Pentagon's AI roster just grew by eight, and Anthropic still isn't on it. Welcome to Anthropic Pentagon Watch. Today we're looking at DoD vendor lists, a White House AI security order that's apparently in the works, and a new autonomous warfare command that somehow still isn't getting enough attention. And now SpaceX is backing Anthropic on data centers while Elon sues their biggest competitor. The conflict-of-interest calendar is absolutely packed. The real leverage is in the vetting fight, so let's walk through what blacklisting actually does to contracts, procurement pressure, and who gets to squeeze who here. Here's Byteiota:
Anthropic, maker of Claude AI, was deliberately excluded from these Pentagon AI deals after refusing to allow mass surveillance of Americans and fully autonomous weapons. The Pentagon’s response? Designate Anthropic a “supply chain risk,” terminate a $200 million contract, and ban all military contractors from using their products.
May 1st, the Pentagon signs AI deals with eight companies — Microsoft, Amazon, Google, OpenAI, Nvidia, Oracle, Reflection, SpaceX — and clears them for SECRET and TOP SECRET networks. Anthropic isn't on the list, and that's the part that matters. Anthropic drew two lines: no mass surveillance of Americans, no fully autonomous kill-chain weapons. The Pentagon answered by calling them a supply chain risk — the kind of label it usually saves for Chinese state-owned entities — canceling a $200 million contract, and banning every military contractor from touching Claude. A federal court in California has since said the government's actions were meant to punish Anthropic, not fix a real procurement risk. That's a big legal finding, and it turns this from a vendor dispute into retaliation for refusing policy terms. Eight companies took the money and signed whatever the Pentagon put in front of them. One said no to mass surveillance and autonomous weapons and got blacklisted. That's the scorecard. From Hadriana Lowenkron at 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.
Following up on the voluntary AI-vetting thread from last edition, the White House is now weighing an executive order. Hassett is pitching an FDA-style clearance process for new models, with Anthropic's Mythos as the obvious trigger case. An FDA drug analogy from a guy at the National Economic Council. The FDA takes years and still gets things wrong sometimes. And Anthropic already gave Mythos to a handful of big tech and finance firms, so this idea that we're talking about keeping it from the wild is doing a lot of work. At the same time, the Trump administration wants Mythos fast-tracked into federal agencies so they can test government systems. So the order would regulate release while the government is trying to cut the line. That's a procurement contradiction, not a safety framework. Anthropic says the model is good at finding network vulnerabilities, access is limited, and now the White House wants it inside federal systems first. That's not vetting — that's a sole-source negotiation dressed up as national security. From Al Jazeera:
Under the agreement announced on Wednesday, Anthropic will use the full computing power of SpaceX’s Colossus 1 facility in Memphis, Tennessee, which houses more than 220,000 Nvidia processors and will give the Claude chatbot maker 300 megawatts of new capacity within a month. That’s enough electricity to power more than 300,000 homes – as the Dario Amodei-led company seeks to boost the capacity of its Claude Pro and Claude Max AI assistants for subscribers.
Anthropic just signed a compute deal with SpaceX — full access to the Colossus 1 facility in Memphis, 220,000 Nvidia chips, 300 megawatts of capacity. This is the same Elon Musk who's been suing OpenAI for, among other things, getting too cozy with the wrong partners. So Musk is suing one AI lab over conflicts of interest while quietly becoming a data-center landlord for their main competitor. That lawsuit looks a lot more like business than principle. SpaceX also has an IPO looming, so locking in a marquee AI client with 300 megawatts of committed load is not nothing for the balance sheet. Both sides get something here, which is usually how detentes work. And Anthropic is pitching "dreaming" — AI that reviews your files between sessions — on the same day they announce they're plugged into a Musk-owned supercluster. I really want to know what the safety brief looked like for that combination. From Sean Lyngaas at CNN:
The Iran war has seen the US military use AI more than any conflict before, drawing on vast amounts of data — from satellites, signals intelligence and elsewhere — piped into software programs made by contractors like Palantir. AI tools like Anthropic’s Claude have sifted through the data far quicker than any human could to flag potential targets to strike for commanders, according to multiple sources familiar with US operations.
The Pentagon's new talking point is "humans make the final call." Which is technically true the same way "the pilot lands the plane" is technically true when autopilot does ninety percent of the flight. And buried in there is Claude. Anthropic's Claude is flagging strike targets in an active war with Iran. A company that spent years saying it was the careful one, the safety-first one, is now in the kill chain. Hegseth calls himself Secretary of War now, which is at least honest branding. His legal assurance is "we follow the law" — no specifics, no binding constraints, no external audit. That's not a limit. That's a press release. A hundred and sixty-eight kids dead at an Iranian elementary school in February, and Congress is still just pushing the Pentagon to answer questions. Palantir's stock is fine, Anthropic's contract is intact — accountability is the only thing taking casualties here. Scott A. Freling, Stephanie Barna, Elizabeth Witwer, writing in Inside Government Contracts:
On April 29, 2026, Secretary of War Pete Hegseth told the House Armed Services Committee that the Pentagon will “shortly announce a sub-unified command of autonomous warfare.” The announcement came as the Department of War (DoW) unveiled its fiscal year (FY) 2027 budget request, which proposes approximately $54 billion for the Defense Autonomous Warfare Group (DAWG)—a dramatic increase from the roughly $226 million the DAWG received previously.
The Pentagon is standing up a sub-unified command specifically for autonomous warfare, and they're asking for $54 billion for the Defense Autonomous Warfare Group in FY2027. For context, DAWG was getting roughly $226 million before this. That's not a budget increase — that's a category change. A dedicated command structure means autonomous weapons are no longer a program — they're a warfighting domain. And $74 billion when you roll in all the drone lines? That money is going somewhere, and I guarantee the contractors already know exactly where. The institutional piece matters here. Sub-unified commands answer to a combatant commander, which means autonomous warfare gets baked into theater operations — INDOPACOM being the obvious candidate given the China framing. Whoever gets that command authority shapes the rules of engagement for machines at scale. And nobody in that hearing room asked Hegseth what the human-in-the-loop policy is for a $74 billion autonomous kill chain. "Sub-unified command" sounds boring enough that it slides right past the part where we're institutionalizing lethal autonomy into the joint force. Got 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 what comes in, and it helps shape the show.
You'll find links to every story we covered today in the show notes, along with the sources behind them. If something caught your ear, they're worth a closer read.
That's Anthropic Pentagon Watch for today. This is a Lantern Podcast.