Google just made a classified AI deal with the Pentagon. Employees are furious.

Google has reportedly signed a classified deal with the US Department of Defense, giving the Pentagon access to its AI models for “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 using the company’s AI over fears it would be used in “inhumane or extremely harmful ways.”

If confirmed, this puts Google in the same club as OpenAI and xAI, both of which have already signed classified AI contracts with the US government. Anthropic was also in that group until it got blacklisted by the Pentagon for refusing to remove safeguards the DoD found inconvenient. That part is worth remembering.

The timing here is brutal. Google employees have been vocal about military AI before. Back in 2018, thousands of them forced the company to drop Project Maven, a Pentagon drone imagery analysis program. That was a rare win for internal dissent. Now, seven years later, we’re back to basically the same fight, except the stakes are higher and the deal is classified, meaning employees can’t even see what they’re being asked to build.

Let’s be clear about what “any lawful government purpose” means in practice. It’s a deliberately broad phrase. It covers everything from logistics optimization and cybersecurity to targeting assistance and autonomous systems. The Pentagon’s lawyers will decide what’s lawful, not Google’s ethics board. And if history is any guide, the definition of “lawful” tends to stretch when national security is on the line.

Google’s official position has always been that it follows the law and applies its AI Principles. Those principles, released after the Maven backlash, explicitly prohibit AI for weapons or surveillance that violates international norms. But a classified deal with the Pentagon is a test of whether those principles are real or just PR. The fact that the deal is classified means we’ll never know the full scope unless someone leaks it.

The contrast with Anthropic is telling. Anthropic got blacklisted because it refused to remove safety guardrails from its models. That’s a company that actually stuck to its principles and paid a price. Google, meanwhile, is reportedly doing the opposite. It’s hard not to see this as a strategic decision: keep the Pentagon happy, keep the contracts flowing, and hope the employees don’t revolt again.

But the employees are already revolting. The internal letter demanding Pichai block the deal was signed by a significant number of staff, and it’s unlikely to die down quietly. Google’s leadership has a choice: listen to their workforce or prioritize the Pentagon’s money. Given the track record, I’m not optimistic.

This deal also raises a broader question about the AI industry’s relationship with the military. OpenAI and xAI are already in, and now Google is joining them. Microsoft has been selling AI to the Pentagon for years through Azure. Amazon’s Rekognition was used by law enforcement. The pattern is clear: the AI giants are lining up to serve the US defense establishment, and the only companies holding out are either too small to matter or, like Anthropic, willing to lose business to keep their principles.

I don’t think AI in defense is inherently evil. There are legitimate uses like logistics, medical triage, or threat detection that save lives. But the problem is the lack of transparency. Classified deals mean no oversight, no public debate, and no accountability. If a Google model is used in a drone strike that kills civilians, we’ll never know. The company will say it followed the law, and the Pentagon will say it was lawful, and that’s the end of it.

The irony is that Google’s own AI Principles were supposed to prevent exactly this kind of slippery slope. They said the company would not design or deploy AI for weapons. But a classified deal with the Pentagon is a loophole big enough to drive a tank through. “Any lawful government purpose” could easily include weapons, as long as the lawyers sign off.

I’ve been watching this space for over a decade, and I have to say: this feels like a regression. The Maven protests were a high-water mark for employee activism in tech. Now, Google is quietly walking back those commitments while the Pentagon gets access to some of the most powerful AI models on the planet. The employees are right to be furious.

What happens next? The deal is reportedly already signed, so reversing it would take a major internal revolt or a public scandal. The employees are organizing, but Google’s leadership has shown it can weather protests. The real question is whether the broader public cares. If this story fades in a week, Google gets what it wants. If it sticks, we might see another round of resignations and policy reviews.

Either way, the precedent is set. Classified AI deals are the new normal, and the tech industry’s ethical boundaries are being redrawn behind closed doors. That’s not a future I’m comfortable with.