The U.S. Department of Defense has signed a contract to use Google’s artificial intelligence models, The Information reported April 28, citing an unnamed source.
Under the agreement, the Pentagon may use Google’s AI models for “any lawful government purpose,” including classified tasks such as operational planning and weapons targeting. The contract reportedly stipulates that both parties agree the AI system is not intended — and must not be used — for large-scale domestic surveillance or for autonomous lethal weapons (including target selection) without proper human oversight and control. At the same time, the agreement makes clear it does not give Google any authority to control or veto the government’s lawful operational decisions.
Analysts say that language could be interpreted to allow the Pentagon, in certain circumstances, to employ Google’s Gemini model for U.S. surveillance or autonomous weapons. A Google spokesperson told Reuters, “Providing API access to our commercial models, including Google infrastructure, under industry-standard practices and terms, is a responsible way to support national security.”
Reuters reported Google is the third company to sign an AI deal with the Pentagon, following OpenAI and Elon Musk–backed xAI. Last year, the Pentagon signed AI-use contracts of up to $200 million apiece (approximately 295 billion KRW) with several major AI firms, including Google, OpenAI and Anthropic.
Anthropic pushed back after the Pentagon’s broad “any lawful government purpose” language, saying it would not permit uses for “large-scale domestic surveillance” or the “development of autonomous lethal weapons.” The Pentagon subsequently terminated its agreement for Anthropic’s Claude model, and the Trump administration issued a ban on Claude’s use within U.S. government agencies.
After President Trump recently said the Pentagon might be allowed to use Claude, observers say the department could reassess and potentially enter a new contract.
