Google quietly inks a classified AI deal with the Pentagon — employees are not happy

Google has reportedly signed a classified deal with the US Department of Defense, allowing the Pentagon to use its AI models for what the agreement vaguely calls “any lawful government purpose.” The Information broke the story, and it landed less than a day after a group of Google employees publicly demanded CEO Sundar Pichai block the Pentagon from accessing the company’s AI, citing fears it would be used in “inhumane or extremely harmful ways.”

If this deal is real — and it sounds real — Google joins a club that already includes OpenAI and xAI, both of which have inked their own classified agreements with the US government. Anthropic was also in that group until recently, but the Pentagon reportedly blacklisted it after the company refused to remove safety restrictions the DoD found inconvenient. So much for responsible AI.

This isn’t Google’s first rodeo with military AI. Remember Project Maven? That was the 2018 initiative where Google provided AI to help analyze drone footage. Employees revolted, thousands signed petitions, and Google eventually backed away, issuing a set of AI principles that explicitly banned the use of its technology for weapons or surveillance that violates international norms. Those principles look a little thin right now.

The timing here is particularly awkward. Google has been trying to position itself as the responsible, cautious player in the AI arms race. They talk about safety, they publish research on alignment, they even have an entire division dedicated to “Responsible AI.” But a classified deal with a vague scope of “any lawful government purpose” — that’s a pretty wide door. What’s considered lawful today might not be tomorrow, and the Pentagon’s interpretation of “lawful” has historically been generous.

I get it: national security contracts are lucrative, and the US government isn’t going to stop wanting AI tools just because some tech workers feel queasy. But Google’s internal culture has always been more activist than, say, OpenAI’s. Employees have a track record of pushing back, and they’ve won before. The question is whether Pichai will cave to pressure again, or whether he’ll double down and argue that this deal is somehow consistent with Google’s AI principles.

Spoiler: it’s probably not. The principles say Google won’t build AI for weapons or surveillance that violates human rights. A classified military deal with no public oversight? That’s exactly the kind of thing those principles were supposed to prevent. Either Google is about to revise its principles, or it’s about to face another internal revolt. Maybe both.