The employees' letter does not argue that Google's AI is inherently dangerous — it argues that classified deployment makes Google's own oversight impossible. On air-gapped networks, the company cannot see what its systems are doing, which means every internal ethics review, every responsible AI policy, and every use-case restriction exists only on paper. The 580-plus signatories pressing Pichai to refuse classified military work are making a procedural point: a guardrail you cannot verify is not a guardrail. Google has proceeded with the Pentagon partnership anyway, which means the company has…