The Pentagon is rolling out Google’s Gemini artificial intelligence agents to its workforce of three million personnel, a move signaling a significant expansion of AI’s role within the Department of Defense. The initial deployment, confirmed by Bloomberg, focuses on unclassified networks and includes eight pre-designed agents intended to automate tasks like summarizing meetings, developing budgets, and evaluating decisions against national defense strategy. Users will also have the ability to build custom agents using natural language, requiring no specialized technical skills.
This large-scale integration of AI comes as the Defense Department seeks to leverage generative AI for increased efficiency and strategic advantage. However, a critical gap exists between deployment and preparedness: only 26,000 employees have completed official training on how to effectively and safely utilize these new AI tools, raising concerns about potential errors and the need for human oversight. The rollout of Gemini agents represents a substantial investment in AI, but its success hinges on addressing this training disparity.
Gemini’s Integration and the GenAI.mil Platform
Emil Michael, Under Secretary of Defense for Research and Engineering, described Google as a “trusted” partner and confirmed ongoing negotiations to extend Gemini’s access to classified networks. Michael, who previously led contentious negotiations with Anthropic – a company recently designated as a supply chain risk – indicated the Pentagon is “moving on” from that dispute, despite Anthropic’s legal challenge. The deployment is being facilitated through GenAI.mil, the Pentagon’s platform for generative AI, which launched in December 2025. Since its launch, employees have submitted 40 million prompts and uploaded over four million documents.
The sheer volume of interaction with GenAI.mil demonstrates a rapid adoption of AI within the Department of Defense, even in a highly regulated environment. However, the limited training – less than 1% of the workforce has received formal instruction – underscores a potential vulnerability. Michael acknowledged the importance of training to mitigate “hallucinations” – instances where AI generates inaccurate or misleading information – emphasizing the need for human review of AI-generated outputs. “It saves you a lot of time in the process, but you have to review at the end to make sure there are no hallucinations,” he stated.
Accelerated Workflows and a Return to Collaboration
The potential for AI to accelerate complex tasks is already becoming apparent. Kenneth Harvey, director of the Mission Training Complex at Fort Bragg, explained that designing a military exercise simulating 50,000 soldiers previously took his team of nine people six months. Using the GenAI.mil portal, a comparable exercise focused on the Southern Command was completed in just six weeks. Harvey stressed that “human eyes reviewed every word,” but the example illustrates the speedup the Pentagon hopes to achieve.
This deployment marks a significant shift for Google, which faced internal backlash in 2018 over its involvement in Project Maven, an AI program for drone imagery analysis. More than 4,000 Google employees signed a protest letter, leading the company to discontinue the contract and establish principles for responsible AI development. Now, eight years later, Google is re-engaging with the Pentagon with a higher profile and, notably, without similar public protests. Google’s Vice President Jim Kelly described the agreement as a collaboration to “automate routine tasks” in a company blog post.
Filling the Gap Left by Anthropic
The Pentagon’s push to integrate AI is occurring rapidly as it rebuilds its supplier ecosystem. Following a disruption with Anthropic – whose Claude system was the only AI model with Impact Level 6 certification for fully classified networks – the Department of Defense has accelerated agreements with OpenAI, xAI, and now Google to cover various network layers. The loss of Claude’s capabilities presents a challenge, as none of its replacements currently possess the same level of security clearance.
The GenAI.mil portal will continue to offer the Gemini chatbot for unclassified work, and now also the new AI agents. Michael expressed “high confidence” that Google will eventually become a partner for classified and top-secret networks, though no specific timeline has been established. The Pentagon’s embrace of Gemini signals a broader trend of increasing reliance on commercial AI solutions for national security purposes.
As the Department of Defense continues to integrate Gemini and other AI tools, the focus will likely shift towards expanding training programs and establishing robust safeguards to ensure responsible and accurate AI implementation. The success of this initiative will depend on bridging the gap between technological deployment and workforce preparedness.
What are your thoughts on the Pentagon’s adoption of AI? Share your comments below.
