The Pentagon will use Gemini for administrative tasks in unclassified networks, with the possibility of extending to classified information.

The Pentagon has decided to implement the artificial intelligence developed by Google, based on the Gemini language model, to assist over three million civilian and military employees in administrative tasks.

Initially, the AI will operate only in unclassified networks, but there are discussions about expanding its use in networks with classified information. Eight AI agents will be used for activities such as summarizing meetings, building budgets, and evaluating national security proposals. Pentagon employees began testing the AI in December, already having over 40 million unique commands and processing more than 4 million documents.

Despite some internal opposition, the Pentagon seeks to diversify its artificial intelligence solutions, following disagreements with Anthropic and the agreement with OpenAI.