Can the Pentagon's AI Ethics Debate Deter Startups From Entering the Defense Sector?

In the span of barely ten days, a high-stakes negotiation between the Pentagon and Anthropic collapsed, the Trump administration formally classified Anthropic as a supply-chain risk, and the AI company announced it would challenge that designation through the courts — a sequence of events that has sent shockwaves through the defense-tech sector.

OpenAI moved swiftly to fill the vacuum, announcing its own agreement with the Department of Defense — a decision that triggered immediate and vocal backlash. The fallout was measurable: ChatGPT uninstalls surged, Anthropic's Claude climbed to the top of the App Store charts, and at least one senior OpenAI executive resigned, citing concerns that the deal was rushed through without adequate ethical guardrails.

On the latest episode of TechCrunch's Equity podcast, Kirsten Korosec, Sean O'Kane, and I unpacked what this escalating standoff means for the broader landscape of AI startups pursuing federal contracts — particularly those eyeing Pentagon partnerships. As Kirsten put it, "Are we going to see a changing of the tune a little bit?"

Sean noted that this situation is unusual on multiple fronts, not least because OpenAI and Anthropic produce platforms that "no one can shut up about." More critically, the core dispute centers on "how their technologies are being used or not being used to kill people" — a framing that almost guarantees a level of public and political scrutiny that most defense contractors never face.

Even so, Kirsten argued, the episode should "give any startup pause."

Read a preview of our conversation, edited for length and clarity, below.

Kirsten: I'm wondering if other startups are starting to look at what's happened with the federal government, specifically the Pentagon and Anthropic, that debate and wrestling match, and [take] pause about whether they want to be going after federal dollars. Are we going to see a changing of the tune a little bit?

Sean: I wonder about that too. In the near term, I'd lean toward no — at least for most players. When you survey the full spectrum of companies, from early-stage startups to established Fortune 500s, that hold active defense contracts, the vast majority of that work operates entirely below the public radar.

General Motors has supplied defense vehicles to the Army for decades, including work on electrified and autonomous variants, and it barely registers in the news cycle. That kind of quiet, sustained engagement with the defense industrial base is the norm, not the exception. What OpenAI and Anthropic ran into this past week is a fundamentally different dynamic — these are companies whose products are embedded in the daily lives of millions of consumers, and more to the point, platforms that no one can stop talking about.

That consumer visibility creates an intensity of scrutiny that most defense contractors simply don't face. The spotlight on these AI labs amplifies their Pentagon involvement to a degree that traditional defense primes — or even most dual-use tech firms — never have to contend with.

There's one important caveat, though. A significant portion of the heat surrounding the Anthropic-OpenAI-Pentagon triangle stems from something more visceral than brand recognition — it's the direct question of whether these AI systems are being deployed in lethal decision-making contexts. That's a categorically different conversation than debating General Motors' role as a defense contractor. The ethical weight is more immediate, more legible to the public, and harder to abstract away.

I don't expect companies like Applied Intuition or others that have deliberately positioned themselves as dual-use platforms to pull back meaningfully. The public scrutiny just isn't there at the same level, and there's no comparable shared cultural understanding of what the downstream impact might look like.

Anthony: What makes this story so compelling — and so difficult to generalize from — is how deeply it's shaped by the specific companies and personalities involved. There's been a wave of substantive commentary exploring the broader questions: What is the appropriate role of technology in government? Of AI in national security infrastructure? Those are legitimate and necessary conversations.

But the lens here is a peculiar one, because Anthropic and OpenAI are not as ideologically divergent as the current narrative suggests. This isn't a story of one company refusing government work while the other embraces it unconditionally, or one demanding strict use-case restrictions while the other offers a blank check. Both organizations, at least in their public posture, have stated they want meaningful constraints on how their AI is deployed. The distinction appears to be one of degree and resolve — Anthropic seems to be drawing a harder line on the specific contractual terms it's willing to accept.

And layered on top of the policy disagreement is what appears to be a genuine interpersonal dimension. Anthropic's CEO and Emil Michael — whom many TechCrunch readers will recognize from his tenure at Uber and who now serves as Chief Technology Officer for the Department of Defense — reportedly have a fraught personal dynamic that may be coloring the entire negotiation. Reportedly.

Sean: Yes, there's a very significant interpersonal conflict element here that we shouldn't gloss over.

Kirsten: Absolutely, though the stakes extend well beyond any personal rivalry. To step back and look at the structural picture: what we're witnessing is a breakdown between the Pentagon and Anthropic in which Anthropic appears to have come out on the losing end of the immediate negotiation — even as its technology remains actively deployed by the military and is considered a critical capability. OpenAI has effectively stepped into that opening, though the situation remains fluid and will almost certainly have evolved further by the time this episode publishes.

The fallout has been telling for OpenAI. ChatGPT uninstalls surged 295% in the wake of the company's decision to formalize its partnership with the Department of Defense — a sharp signal that a meaningful segment of its user base views military entanglement as a line crossed.

But the uninstall numbers are a distraction from the more consequential issue at play. The Pentagon wasn't just negotiating a new agreement — it was actively seeking to modify the terms of an existing, already-executed contract. That distinction matters enormously, and it should put every startup eyeing federal procurement on notice. Government contracting has always been a slow, bureaucratic process by design; terms get hammered out over months, sometimes years, before ink hits paper. The fact that the DoD is now moving to retroactively reshape those terms suggests the current political environment around defense AI is operating by a different set of rules — and that the usual assumptions about contract stability may no longer hold.