Google steps in after Anthropic tells the Pentagon no

Anthropic made a principled stand this month by refusing to let the Department of Defense use its AI models for domestic mass surveillance and autonomous weapons. Google, unsurprisingly, had no such qualms.

The search giant has now signed a new contract with the DoD, effectively filling the gap Anthropic left behind. This isn’t Google’s first military rodeo — remember Project Maven? That 2018 deal, which used AI to analyze drone footage, sparked massive internal protests and ultimately led to Google not renewing the contract. But times change, or maybe just the public relations calculus.

Anthropic’s decision was framed around its responsible AI principles, specifically prohibiting uses that could lead to harm or violate human rights. And look, I get it — the company wants to be the good guy in the AI safety narrative. But here’s the thing: when you refuse a government contract, someone else will take it. In this case, that someone is Google, which has been aggressively rebuilding its defense relationships over the past few years.

Google’s new contract reportedly covers a broader scope than what Anthropic was asked to provide, including access to Gemini models for various defense applications. The details are predictably vague — these are government contracts, after all — but the pattern is clear: if you won’t do it, Google will.

I find this whole situation a bit frustrating. Anthropic gets to look clean and principled, while Google takes the money and the heat. But is refusing a contract really a moral victory if the same work gets done anyway? It feels more like passing the buck than making a real difference.

What’s more interesting to me is how Google’s internal culture has shifted since the Maven days. Back then, thousands of employees signed petitions and some even resigned. Now? Crickets. Either the workforce has been sanitized, or people have just accepted that this is how the AI arms race works.

The Pentagon isn’t going to stop wanting AI capabilities, regardless of which company provides them. The real question is whether companies like Anthropic can push for meaningful restrictions in their contracts — not just refuse to participate. That would actually change something.