Google Military AI Protest Reenacts: Over 600 Employees Sign Petition Asking Pichai to Reject Pentagon Confidential Contracts

According to Beating Monitoring, over 600 Google employees jointly wrote to CEO Sundar Pichai, demanding that the company refuse to deploy AI models on the Pentagon’s classified networks. The organizers stated that many signatories are Google DeepMind researchers, as well as more than 20 directors and vice presidents at the executive level.

The letter states: “The only way to ensure Google is not associated with such harm is to refuse all confidential workloads. Otherwise, these uses may occur without our knowledge and without our ability to prevent them.” The employees’ core concern is that: the classified networks are air-gapped from the public internet, and Google cannot monitor the actual use of the models or implement any technical safeguards.

The direct trigger for the joint letter was a report by The Information earlier: Google is negotiating with the Pentagon to deploy Gemini in classified environments, with contract terms allowing the Pentagon “for all lawful purposes.” Currently, Google has provided the Gemini service to 3 million personnel via the genAI.mil platform for non-classified use. Deputy Secretary of Defense for Research and Engineering Emil Michael told Bloomberg in March that non-classified use is just the beginning, “next, it will move into classified and top-secret levels.”

This marks Google’s second large-scale internal protest regarding AI militarization. In 2018, over 4,000 employees jointly opposed Project Maven (using AI to analyze drone footage for the Pentagon), dozens resigned, and Google ultimately declined to renew the contract, establishing principles against using AI for weapons and surveillance. However, in February 2025, Google quietly removed these ban clauses, with DeepMind CEO Demis Hassabis stating on the same day that “democratic nations should lead in AI development and support national security.” Since then, Google has secured part of the $9 billion JWCC military cloud contract. Meanwhile, Anthropic has been designated as a “supply chain risk” and faces legal action by the Pentagon for refusing to relax restrictions on the use of Claude by the military.