Pentagon-Anthropic rift could deter startups from defense work

Within just over a week, negotiations over how the Pentagon could use Anthropic’s Claude technology fell apart, the Trump administration labeled Anthropic a supply-chain risk, and the AI company vowed to challenge that classification in court.

Meanwhile, OpenAI quickly sealed its own deal, triggering backlash that led users to uninstall ChatGPT and pushed Anthropic’s Claude to the top of the App Store charts. At least one OpenAI executive resigned, citing concerns that the announcement was rushed without proper safeguards.

On the latest episode of TechCrunch’s Equity podcast, Kirsten Korosec, Sean O’Kane, and I discussed what this means for other startups looking to partner with the federal government, especially the Pentagon. Kirsten asked, “Are we going to see a shift in the tune?”

Sean noted that the situation is unusual in several ways, partly because OpenAI and Claude make products that “no one can shut up about.” Crucially, the dispute centers on “how their technologies are being used — or not used — to kill people,” which naturally invites more scrutiny.

Still, Kirsten argued that this scenario “should give any startup pause.”

Below is a preview of our conversation, edited for length and clarity.

Kirsten: I’m curious whether other startups are watching what happened with the federal government — specifically the Pentagon and Anthropic, that debate and wrestling match — and starting to hesitate about pursuing federal dollars. Will we see a shift in the tune?

Sean: I wonder the same. In the near term, I think probably not, if only because when you look at the many companies — whether startups or more established Fortune 500s — that work with the government, especially the Department of Defense or the Pentagon, a lot of that work flies under the radar.

General Motors has built defense vehicles for the Army for a very long time and has worked on all-electric and autonomous versions. That kind of work happens constantly and rarely enters the public consciousness. The problem OpenAI and Anthropic ran into last week is that these are companies making products used by a ton of people — and, more importantly, products “no one can shut up about.”

So there’s a huge spotlight on them, which naturally highlights their involvement to a degree that most other companies contracting with the federal government — especially its war-fighting branches — don’t have to deal with.

The one caveat I’d add is that much of the heat in the discussion between Anthropic, OpenAI, and the Pentagon is very specifically about how their technologies are or aren’t being used to kill people, or in parts of missions that do. It’s not just the attention and brand familiarity; there’s an extra layer that feels more abstract when you think about General Motors as a defense contractor.

I don’t think we’ll see Applied Intuition or other dual-use companies pull back much, simply because the spotlight isn’t on them, and there’s no shared understanding of what that impact might be.

Anthony: This story is really unique and very specific to these companies and personalities. There have been plenty of thought pieces about the role of technology — and AI — in government, and those are all worthwhile questions to explore.

That said, this is a curious lens through which to examine those ideas because Anthropic and OpenAI aren’t actually that different in many ways, or in the stances they’re taking. It’s not that one company says “I don’t want to work with the government” and the other says “Yes, I do.” Or that one says “do whatever you want” and the other says “I want restrictions.” Both, at least publicly, say they want restrictions on how their AI is used. It just seems Anthropic is digging in its heels more about not letting the terms be changed.

On top of that, there seems to be a personality layer: the CEO of Anthropic and Emil Michael — whom many TechCrunch readers may remember from his Uber days, now chief technology officer for the Department of Defense — apparently really don’t like each other, reportedly.

Sean: Yes, there’s a very big “girls are fighting” element here that we shouldn’t overlook.

Kirsten: Yeah, a little bit. But the implications go beyond that. Pulling back a bit, what we’re talking about is the Pentagon and Anthropic coming to a dispute where Anthropic appears to have lost — though I should note the military is still using their technology. It’s considered crucial. But OpenAI stepped in, and this is evolving; it will likely change by the time this episode airs.

The blowback for OpenAI has been interesting: ChatGPT uninstalls surged about 295% after the deal with the Department of Defense was locked in.

To me, all this is noise around the really critical and dangerous thing: the Pentagon was seeking to change existing terms on an existing contract. That is significant and should give any startup pause, because the political machine at the DoD right now seems different. This isn’t normal. Contracts at the government level take forever to bake in, and the fact they’re trying to change those terms is a problem.

Loading the player…