Google quietly picks up the Pentagon contract Anthropic turned down

Anthropic made a principled stand last week. The company refused to let the Department of Defense use its Claude models for domestic mass surveillance or autonomous weapons. It was a clear line in the sand, and frankly, it got a lot of people nodding their heads.

Now Google has signed a new contract with the Pentagon. The details are sparse, but the timing is not subtle. This is the same kind of work Anthropic walked away from.

Let me be clear: I’m not saying Google is doing something illegal. But the optics here are terrible. A company that built its brand around “don’t be evil” is now the backup plan for one of the most controversial military AI applications out there.

Google’s official line is that this contract falls under their existing work with the DoD on “responsible AI” and cybersecurity. They’ve trotted out the usual talking points about keeping Americans safe and adhering to ethical guidelines. But we’ve heard this before. Remember Project Maven? Google employees revolted in 2018 over drone strike analysis, and the company backed down. Now they’re back at the table.

What changed? My guess is the competitive pressure. Microsoft and Amazon have been eating Google’s lunch in the government cloud space for years. The Pentagon’s Joint Warfighting Cloud Capability contract alone is worth billions. Google can’t afford to keep playing the moral high ground if it wants a seat at that table.

Anthropic’s refusal set a new benchmark. By saying no publicly, they forced every other AI company to answer the same question: where’s your line? Google’s answer, apparently, is somewhere past mass surveillance and autonomous weapons.

I don’t think this is about malice. I think it’s about institutional inertia. Once you’re inside the Pentagon’s procurement machine, the default answer to “can you do this?” is always yes. The ethical review becomes a checkbox exercise.

The DoD isn’t going to stop wanting these capabilities. If Google hadn’t taken the contract, Oracle or Palantir would have. But that’s not really the point. The point is that Google had a chance to draw the same line Anthropic drew, and they chose not to.

For now, the contract details are classified. But we know enough to be concerned. Domestic mass surveillance is not hypothetical. The Pentagon has been pushing for expanded surveillance powers for years, and AI makes it cheaper and more effective. Autonomous weapons are already being tested in various forms. Giving Google’s AI access to those systems is a bet I’m not sure we should be making.

I’ll be watching for employee reactions. Google’s workforce has a history of pushing back on military contracts. If the internal protests start, we’ll know this goes deeper than the press release suggests.