Google Steps In After Anthropic Tells the Pentagon No

Anthropic made headlines recently for telling the Department of Defense it wouldn’t let its AI models be used for domestic mass surveillance or autonomous weapons. That’s a principled stand, and one that got a lot of applause from people who worry about where this technology is heading.

Google, meanwhile, went the other direction. The company just signed a fresh contract with the Pentagon, effectively stepping in where Anthropic refused to go. This isn’t some small pilot program either. It’s a direct continuation of Project Maven-style work, with Google offering its AI capabilities for military applications that include surveillance and weapons systems.

I’ve been watching this space for a while, and the timing is telling. Anthropic’s refusal was public and loud. Google’s move feels quieter, more business-as-usual. But the contrast is stark. One company draws a line on ethical grounds, and another happily crosses it for a government contract.

To be fair, Google has been down this road before. Remember the employee backlash over Project Maven back in 2018? Thousands of staff signed a petition, and Google eventually backed away from renewing the contract. The company even published a set of AI principles that specifically ruled out using AI for weapons or surveillance. Those principles are still on their website. They just don’t seem to apply anymore.

This new deal isn’t exactly the same as Maven, but it’s close enough that the distinction feels academic. The Pentagon wants AI for analyzing drone footage, identifying targets, and potentially automating decisions about who gets attacked. Anthropic said no to that. Google said yes.

What bothers me most is the lack of transparency. Google didn’t announce this with a blog post or a press release. It leaked out through procurement documents and industry chatter. The company knows this is controversial, so it’s trying to fly under the radar. That’s not how ethical leadership works.

Anthropic’s decision wasn’t cost-free. The Pentagon is a massive customer, and walking away from that revenue takes guts. But it also sends a signal that not every AI company is willing to trade principles for profit. Google’s decision sends the opposite signal.

I’m not naive about defense contracting. The US military has legitimate needs, and technology can save lives in the right context. But domestic mass surveillance and autonomous weapons are a different category. They’re the kind of applications that civil liberties groups and even some military ethicists have warned about for years.

Google’s new contract shows that corporate AI ethics are still mostly performative. When the money is big enough, the principles get quietly shelved. That’s a shame, because Google has some of the best AI researchers in the world. They could be leading on responsible AI. Instead, they’re proving that Anthropic’s caution was justified.

The Pentagon will get its AI tools one way or another. The question is which companies are willing to provide them, and what that says about the industry’s moral compass. Right now, the answer isn’t comforting.