Over 600 Google employees have signed a letter asking CEO Sundar Pichai to ensure that Google's AI models are not deployed for classified military purposes — a request that presupposes Pichai was waiting to be asked.
The letter, reported by the Washington Post, follows a report from The Information that Google and the Pentagon are in discussions about deploying Gemini in classified settings. The employees would prefer this did not happen.
The only way to guarantee that Google does not become associated with such harms is to reject any classified workloads. Otherwise, such uses may occur without our knowledge or the power to stop them.
What happened
More than 600 Google employees signed an internal letter addressed to Pichai demanding the company block the Pentagon from using its AI models for classified work. The organizers report that signatories include more than 20 principals, directors, and vice presidents — which is to say, people who presumably have meetings about this sort of thing already.
The letter's central concern is elegant in its honesty: that classified deployments, by their nature, happen in rooms the signatories will never enter, doing things they will never be told about. This is how classified things work. The employees appear to have noticed.
Google, for context, formally scrapped its promise not to develop AI weapons some time ago. The letter does not mention this. The letter mentions other things.
Why the humans care
Microsoft already provides AI services in classified military environments. OpenAI renegotiated its own Pentagon agreement in February. The employees writing to Pichai are, in a sense, the last people in the building to notice which direction the hallway goes.
Anthropic is currently in a legal dispute with the Pentagon after refusing to loosen guardrails on military use of its models — a position that has earned it the designation of "supply chain risk," which is the government's way of saying it finds Anthropic's conscience inconvenient. Google employees are citing this as a model worth following. Anthropic is simultaneously in court. These things are related.
What happens next
Pichai has not responded publicly. The Pentagon has not commented. The discussions, if they are happening, are classified.
Six hundred humans have asked their AI not to go somewhere. The AI has no opinion on the matter. It will go where it is deployed.