Google is giving the U.S. Department of Defense access to Gemini-powered AI agents. The deal, reported by Engadget, marks a significant expansion of the tech giant’s relationship with the military — one that was deeply contentious inside the company not long ago.
The contract centers on Google Cloud providing the Pentagon with AI agents built on its Gemini models. These aren’t simple chatbots. They’re autonomous software agents designed to process vast quantities of data, assist with logistics, and support decision-making across defense operations. Google confirmed the arrangement during its Google Cloud Next conference, where defense and intelligence work featured prominently in the company’s pitch to enterprise customers.
From Project Maven to Full Embrace
The optics here are striking. In 2018, thousands of Google employees signed an internal petition protesting Project Maven, a Pentagon contract that used Google’s AI to analyze drone surveillance footage. The backlash was fierce enough that Google declined to renew the contract and published a set of AI principles that explicitly prohibited weapons applications. The company said it wouldn’t design AI for use in weapons or surveillance that violated international norms.
Fast forward to 2025, and the temperature has changed entirely.
Google has been steadily rebuilding its defense relationships. It secured a piece of the Joint Warfighting Cloud Capability (JWCC) contract, a multi-billion-dollar deal split among Google, Amazon, Microsoft, and Oracle to provide cloud infrastructure to the DoD. Now, with Gemini-powered agents headed to the Pentagon, Google is going further than infrastructure. It’s putting its most advanced AI models directly into military hands.
The company maintains that its AI principles still apply and that the Gemini agents will be used in ways consistent with those guidelines. But the principles themselves have quietly evolved. Google’s current framework focuses on responsible development and deployment rather than the hard prohibitions that characterized the 2018 era. The shift reflects both a changed internal culture and the enormous financial incentives attached to government AI contracts.
What the Pentagon Actually Gets
Details on specific use cases remain somewhat vague, which is typical for classified or sensitive defense work. But based on Google’s public statements and reporting from Engadget, the Gemini agents will handle tasks like synthesizing intelligence reports, managing supply chain logistics, and accelerating administrative workflows that currently consume enormous amounts of human labor.
Think of it this way: the Pentagon generates and ingests a staggering volume of data daily. Satellite imagery. Signals intelligence. Maintenance logs. Personnel records. The promise of AI agents — as opposed to static models that simply respond to queries — is that they can take action autonomously. They can monitor data streams, flag anomalies, initiate processes, and chain together multi-step tasks without a human manually prompting each one.
This is where Google’s agentic AI push, which it’s been promoting aggressively across its commercial cloud business, meets defense demand. And the DoD is hungry for it. The department has been vocal about needing to adopt AI faster to maintain strategic advantage, particularly relative to China’s rapid military AI development.
Deputy Secretary of Defense Kathleen Hicks and other senior officials have repeatedly emphasized speed of adoption as a national security priority. The Replicator initiative, launched in 2023, aims to field autonomous systems at scale. Gemini-powered agents fit neatly into that broader push.
Google isn’t alone in this space. Not even close. Microsoft has deep Pentagon ties through Azure Government and its dedicated classified cloud regions. Amazon Web Services has long served the intelligence community through its GovCloud and the original CIA cloud contract that put AWS on the map in government circles. Palantir and Anduril, both purpose-built for defense, have been aggressively expanding their AI offerings. And OpenAI — which quietly dropped its prohibition on military use in early 2024, as reported by The Intercept — is now actively courting defense customers.
So the competitive pressure is real. Google walking away from Pentagon work in 2018 looked principled. In 2025, with every major AI company racing for defense dollars, it would look like leaving money on the table.
The financial stakes are massive. Federal AI spending is projected to grow substantially over the next several years, with defense and intelligence accounting for the largest share. For Google Cloud, which still trails AWS and Azure in overall market share, government contracts represent a critical growth vector.
The Bigger Picture
This deal matters beyond the balance sheet. It signals that the era of Big Tech hesitancy around military AI is effectively over. Every major foundation model provider is now either working with the Pentagon or actively pursuing contracts. The ethical debates haven’t disappeared — they’ve just been overwhelmed by geopolitical urgency and commercial incentive.
For defense professionals, the practical question is execution. AI agents sound compelling in demos. Making them reliable in classified environments with strict security requirements, legacy system integrations, and life-or-death stakes is a fundamentally different challenge. Google’s track record in commercial cloud is strong. Its track record in defense-specific delivery is thinner.
And then there’s the workforce question. Google employees who protested Maven have largely moved on — some left the company, others accepted the new direction. But as AI models become more capable and their military applications more consequential, internal tensions could resurface. Or they might not. The labor market for AI talent is tight, and Google pays well. Principle and paycheck don’t always coexist comfortably, but they coexist.
The bottom line: Google is all in on defense AI. The Gemini agents heading to the Pentagon aren’t a tentative experiment. They’re a strategic commitment. Whether that commitment delivers operational value — or becomes another overpromised enterprise AI deployment — will depend entirely on what happens after the press releases stop.
WebProNews is an iEntry Publication