The letter that landed in Sundar Pichai's inbox last week wasn't really about one contract. It was an accusation.
Over 600 Google employees—including more than 20 executives from DeepMind—signed a document that amounts to an ultimatum: reject classified military AI workloads, or admit that the company's vaunted AI ethics commitments are corporate theater. "The only way to guarantee that Google does not become associated with such harms is to reject any classified workloads," the letter states. "Otherwise, such uses may occur without our knowledge or the power to stop them."
That final phrase—"without our knowledge or the power to stop them"—is the accusation dressed as a warning. It suggests that Google has learned from its last public mutiny, in 2018, when thousands of employees protested Project Maven and forced the company to walk away from a Pentagon contract. That episode taught leadership something. It taught them to keep these arrangements classified. To give themselves deniability. To maintain the fiction of ethical AI while quietly pursuing the defense dollars that Microsoft and Amazon have already claimed.
The Pentagon wants what Google has. Advanced AI models capable of surveillance analysis, autonomous targeting, and intelligence synthesis. These are the defense contracts of the next decade. If Google doesn't take them, someone else will—and the company has watched its cloud revenue lag behind competitors who made fewer noises about ethics.
But the 600+ signatories aren't buying the trade-off framing. The letter argues that classified workloads aren't a gray area to be managed. They are a line Google must choose to either cross or not. Anthropic, a competitor, is currently in federal court fighting a similar battle with the Pentagon—making this a test case for whether any AI company can successfully resist classified defense work while remaining a prime contractor.
For Pichai, the pressure is now explicit. The signatories include directors, vice presidents, and principal engineers—people who could leave. People whose institutional knowledge is hard to replace. The letter forces a binary choice: commit to blocking classified work, or acknowledge that the company's public commitments to responsible AI development are contingent on what sells.
What happens next is unclear. Google's communications team will likely issue a statement about the company's commitment to ethical AI development while giving Pichai room to maneuver. But the letter's language forecloses that middle path. Either Google makes a binding commitment to refuse classified workloads, or the signatories are telling leadership that they understand exactly what's happening—and they want it on the record.