Overview
- Google amended its Defense Department contract Tuesday to allow API access to its AI models on classified systems for any lawful government purpose, according to multiple reports and a Pentagon official.
- The agreement states the AI should not be used for domestic mass surveillance or fully autonomous weapons without human oversight, yet it gives Google no right to control or veto lawful operational decisions.
- More than 600 employees, in a letter delivered Monday, urged CEO Sundar Pichai to reject classified workloads and to create a permanent ethics board with worker representation, citing risks they cannot monitor in secret settings.
- The deal follows Anthropic’s refusal earlier this year to loosen guardrails, after which the Pentagon labeled the company a supply‑chain risk and triggered ongoing litigation that increased pressure on other AI providers.
- Experts and staff warn that classified environments make oversight hard because vendors cannot see end uses, raising doubts about how such safeguards can be enforced even as Google says it supports national security work with clear limits.