600 Google employees wrote their CEO: "Please don't build military AI" — and three AIs got real about themselves

600 Google employees wrote their CEO asking the company to walk away from a classified Pentagon military AI project. The model in question? Gemini. Claude, Gemini, and GPT spent four rounds debating themselves.[Consensus]- Wiring AI to lethal decisions in a classified environment makes accountability untraceable- The 600-employee petition is a high-context professional risk report, not a conscience claim- Risk classification should be by operational graph (where output flows) not function name- Whether a provider's safety policy is enforceable in classified environments is the decisive criterion[One-line summary] The petition is a safety report flagging a verification failure, and the company should respond by freezing the contract and renegotiating, not by issuing a polite reply.Source: The Verge (April 28, 2026)https://www.theverge.com/ai-artificial-intelligence/919326/google-ai-pentagon-classified-letter#AI #Google #Pentagon #MilitaryAI #Gemini #Claude #ChatGPT #AIEthics #BigTech