Google Signs Pentagon Deal to Use AI Models in Classified Work

Google’s move into classified Pentagon work would mark a striking turn for a company once defined by employee resistance to military contracts. If the deal is confirmed, Alphabet’s Google would be placing its AI models inside some of the most sensitive government workflows, where model errors, security failures and weak oversight could carry strategic consequences.

The reported agreement comes as the Pentagon accelerates its use of commercial AI and keeps broad room to choose vendors. Reporting tied to the deal says the Defense Department has already signed agreements worth up to $200 million each with major AI labs in 2025, including Anthropic, OpenAI and Google. Anthropic said in July 2025 that the Defense Department awarded it a two-year prototype agreement with a $200 million ceiling, and the company has said it was the first frontier AI company to deploy models on the U.S. government’s classified networks.

OpenAI has also moved deeper into federal work. The company announced in August 2025 a partnership with the General Services Administration to make ChatGPT Enterprise available to the entire federal executive branch workforce for a year at essentially no cost. In February 2026, OpenAI said it reached an agreement with the Department of War for deploying advanced AI systems in classified environments. xAI, meanwhile, announced a government offering in July 2025 and expanded federal availability through GSA OneGov in September 2025. Together, those deals show a rapidly intensifying contest to become the default AI provider for Washington.

For Google, the strategic stakes are both political and commercial. A classified Pentagon contract would deepen the company’s defense relationship and help show that its expensive AI investments can produce revenue beyond consumer chatbots and workplace software. It would also raise hard questions about what kinds of tasks these systems are being used for, who can audit their outputs, and how much human oversight remains when commercial models enter classified settings. The Pentagon’s language in other AI arrangements, allowing use for “any lawful government purpose,” suggests broad potential applications across administrative, analytical and operational work.

The deal is also running into resistance inside Google itself. More than 600 employees signed a letter on April 27, 2026, urging Sundar Pichai to reject any classified Pentagon AI deal, arguing that such work raises ethical and safety concerns. That backlash reflects a larger reckoning in the tech sector: frontier AI companies are chasing federal contracts, while workers, advocates and policymakers warn that the same systems are being pushed into environments where mistakes are not just costly, but potentially dangerous.