Google Joins Pentagon’s AI Club – But Not Without Pushback Inside the Company

With input form the New York Times, the Verge, the Guardian, and the Wall Street Journal.

Google is stepping deeper into the defense world. The tech giant has struck a deal with the United States Department of Defense to provide artificial intelligence tools for classified use, putting it alongside rivals already working behind the scenes on sensitive government systems.

The agreement opens the door for Google’s AI models to be used across classified networks for what officials describe as “any lawful government purpose.” That’s a broad mandate. It can cover everything from mission planning to intelligence analysis – areas where AI is quickly becoming indispensable.

Google isn’t alone here. The Pentagon has been building out a roster of AI partners, including OpenAI and xAI, both of which have secured similar contracts to supply models for classified environments. These deals, some worth up to $200 million each, signal how aggressively the military is moving to integrate commercial AI into its operations.

Still, the fine print matters. Google’s agreement reportedly includes limits – at least on paper. The system isn’t meant for domestic mass surveillance or fully autonomous weapons without human oversight. That caveat reflects a line Silicon Valley has long tried to hold, even as government demand pushes the boundaries.

But the same contract also makes clear that Google won’t have veto power over how its tools are used, as long as those uses are lawful. In practice, that leaves a lot of room for interpretation – and that’s exactly what’s making some employees uneasy.

Inside the company, resistance is already bubbling up. More than 600 workers signed a letter urging CEO Sundar Pichai to walk away from classified AI work altogether. Their argument is straightforward: building powerful systems without firm control over how they’re deployed carries real ethical risks.

“We feel that our proximity to this technology creates a responsibility,” the letter said, warning about potential “inhumane or extremely harmful” uses.

It’s not the first time Google staff have drawn a line. Back in 2018, employee protests forced the company to drop its involvement in a Pentagon drone-analysis project known as Project Maven.

The broader context has shifted since then. Google’s parent, Alphabet Inc., quietly rewrote its AI principles last year, removing language that explicitly ruled out work on weapons or surveillance. Executives now frame national security as a legitimate – and even necessary – use case for advanced AI.

That shift helps explain why the Pentagon is leaning harder on commercial players. Government officials have been pushing AI companies to make their tools available on classified networks with fewer restrictions. Not every firm has agreed. Anthropic, for instance, reportedly resisted loosening safeguards, triggering tensions with defense officials and even being flagged as a potential supply-chain risk.

For the Pentagon, the goal is speed and scale. Military planners want access to cutting-edge AI without waiting years for in-house development. For tech companies, the incentives are more complicated – big contracts, yes, but also reputational risk and internal dissent.

Google is trying to thread that needle. A company spokesperson said providing controlled access to its models, with standard safeguards, is a “responsible approach” to supporting national security. Whether that reassurance holds inside the company is another question.

The deal also highlights a bigger trend: the line between Silicon Valley and the defense sector is fading fast. AI is now seen as critical infrastructure, not just another product category. Governments want it. Companies are building it. And employees are increasingly caught in the middle.

For now, Google is moving ahead. The Pentagon gets another major AI partner. And the debate over how far tech firms should go in supporting military applications is only getting louder.