Google Deepens Pentagon AI Partnership with Gemini Agents

Google is expanding its collaboration with the U.S. Department of Defense (DoD), rolling out Gemini-powered AI agents to its more than three million civilian and military personnel. The initiative, launched through the Pentagon’s enterprise AI portal, GenAI.mil, aims to automate routine tasks and enhance efficiency across the department.

AI Agents for Administrative Tasks

The core of this partnership involves providing access to Google’s Gemini AI models, enabling users to build custom AI agents for unclassified work. These agents are designed to assist with repetitive administrative duties, such as drafting meeting notes, creating action items, and breaking down large projects into manageable steps . A no-code or low-code tool called Agent Designer will empower personnel to create these digital assistants without requiring specialized technical expertise .

Expansion to Classified Networks Under Discussion

While the initial deployment focuses on unclassified networks, discussions are already underway to extend the use of these AI agents to classified and top-secret environments . Emil Michael, the DoD’s Under Secretary of Defense for Research and Engineering, expressed “high confidence” in Google’s ability to become a partner across all network levels .

High Adoption Rates and Training Gaps

Since the launch of GenAI.mil in December, DoD employees have submitted 40 million prompts and uploaded over four million documents, demonstrating significant adoption of the AI tools . However, a notable gap exists between usage and training, with only 26,000 personnel completing AI training as of March 2026, despite fully booked future sessions . Michael emphasized the importance of training to mitigate potential errors and identify “hallucinations” – inaccurate or misleading responses generated by AI .

Operational Value Demonstrated

Early results indicate substantial time savings. Kenneth Harvey, director of the Mission Training Complex at Fort Bragg, reported that designing a military exercise with 50,000 simulated soldiers, which previously took his team six months, was completed in just six weeks using the AI portal. While human review was maintained throughout the process, this example highlights the potential for accelerated operations .

A Shift in Google’s Approach

This partnership marks a return to DoD collaboration for Google, following a period of internal debate and protest. In 2018, over 4,000 Google employees signed a letter opposing the Maven Project, which involved applying AI to military drone imagery . Since then, Google has gradually relaxed some of the restrictions established in response to that controversy, reflecting a broader trend within the technology industry towards increased engagement with military contracts .

Navigating a Changing AI Vendor Landscape

The DoD is actively building its AI vendor ecosystem, forging agreements with OpenAI and xAI in addition to Google, following a dispute with Anthropic. Anthropic was designated a “supply chain risk” after refusing to modify its technology to address concerns regarding domestic surveillance and autonomous weapons, and is currently challenging that designation in court . Filling the gap left by Anthropic, whose Claude system was the only frontier AI with the necessary certification for fully classified networks, remains a key challenge .