Google to provide Pentagon with Gemini-powered AI agents

Google said it is adding Agent Designer to GenAI.mil, the Pentagon’s internal generative AI platform, giving civilian and military personnel a way to build their own Gemini-powered agents for unclassified work. The Pentagon is no longer just handing workers a chatbot and asking them to experiment. It is giving them a way to build agents around repeat tasks and fold those agents into daily work.

That is a bigger shift than the original rollout in December. Back then, Google pitched Gemini for Government as the first enterprise AI tool on GenAI.mil for more than 3 million civilian and military personnel, mostly around unclassified productivity work. Now Google says the platform has already drawn more than 1 million unique users in a little over a month.

Why the stakes are higher now

There is a real difference between a chatbot and an agent. A chatbot answers questions. An agent carries out tasks on a user’s behalf. This is important as the Pentagon is moving from general AI access to systems designed to handle repeat work inside daily operations.

The first use cases are administrative. Once personnel start using agents for routine tasks, the infrastructure, trust, and workflow habits are already in place. That makes broader adoption easier later.

Also read: Why OpenAI and Google employees are backing Anthropic in its Pentagon blacklist fight.

Reports suggest Google’s agents are being introduced on unclassified systems, while Emil Michael, the Pentagon’s under secretary of defense for research and engineering, said the the department wants to move next into classified and top-secret environments. So the current rollout is limited, but the direction is clear.

The Anthropic fight changed the context

The launch lands at a very specific moment in the Pentagon’s fight with AI vendors. Anthropic was not pushed aside over some vague policy disagreement. The Pentagon designated the company a supply-chain risk after a clash over guardrails that barred uses tied to domestic surveillance and autonomous weapons. And Google’s expansion is happening in the exact space Anthropic refused to give away.

Seen that way, it is a line-drawing story. OpenAI and Google are moving deeper into defense work at the same time Anthropic is being punished for holding its ground. The Pentagon is making clear what kind of AI partner it wants, and the companies stepping forward are making their choice just as clearly.

What this story is really about

So yes, the official use cases are still framed around unclassified office work. That is the public-facing version of the rollout, and it is real. But the story would be too small if it ended there. Google is becoming harder to describe as just another software vendor with a government contract. The company is positioning itself inside the machinery of defense work itself.

That is the bigger shift underneath the feature update. The Pentagon is not only buying access to frontier AI models. It is starting to build working relationships around agentic systems, delegated tasks, and long-term operational trust. That is how the next generation of defense contractors gets built.