Google, Pentagon Discuss Classified AI Contract, Reuters Says

Google and the U.S. Defense Department are in discussions about a potential classified artificial intelligence deal involving Google’s Gemini technology, Reuters reported, citing The Information.

The talks center on whether Gemini could be used in a classified setting for Pentagon work, according to the report referenced by Reuters. Any arrangement would involve deployment and use of AI in environments where information is restricted for national security reasons.

Google, a unit of Alphabet, has been expanding its enterprise AI offerings as competition among major technology companies intensifies. A Pentagon relationship tied to classified work would place Google’s AI systems closer to core national security operations than many commercial deployments, raising the stakes for reliability, security controls, and compliance requirements.

The Pentagon has been exploring how to apply advanced AI across a range of missions and support functions, and classified use cases typically require strict controls on data handling, access, and infrastructure. A deal involving a widely used commercial AI platform would also underscore the government’s push to work with private-sector developers as the technology moves quickly.

The reported discussions arrive amid heightened attention on how AI should be tested, constrained, and governed when used in sensitive settings. Classified deployments can involve additional scrutiny related to cybersecurity, model behavior, and the management of training and inference data, along with questions about where systems run and who can access outputs.

It was not immediately clear from the report how advanced the discussions are, what specific military or intelligence functions the Pentagon is evaluating, or whether the talks cover a pilot program, a broader contract, or a limited capability demonstration. Reuters’ report did not describe the financial terms or a timeline for a potential agreement.

A Pentagon deal could matter beyond a single contract. Classified deployments often influence standards that can later shape wider government procurement and technical requirements, potentially affecting how AI tools are built and audited for security-sensitive environments.

The next steps would typically include technical evaluations, security reviews, and contracting decisions, along with determinations about how and where any AI capability would operate. The parties could also decide to limit any initial work to narrowly defined use cases while testing performance and safeguards.

For now, the reported talks signal that Google and the Pentagon are considering a path for Gemini to be used in classified work, a move that would deepen the company’s footprint in government AI.