Google is in active negotiations with the Pentagon to deploy its Gemini AI model on classified and top-secret military networks, according to a Bloomberg report published in March 2026. The company is already rolling out Gemini-powered AI agents to Defense Department staff for unclassified tasks, but the classified expansion would move a commercial large language model into environments that handle intelligence, operational planning, and other sensitive national security material.
If the deal goes through, it would represent one of the most significant integrations of commercial AI into the U.S. military’s most guarded systems, and it would place Google at the leading edge of a Pentagon-wide push to embed frontier AI across defense operations.
Google’s existing foothold
Google’s path into the Pentagon’s AI ecosystem has been building for months. In July 2025, the Defense Department’s Chief Digital and Artificial Intelligence Office awarded contracts worth up to $200 million to a group of AI companies that included Google, xAI, OpenAI, and Anthropic, as Bloomberg reported at the time. Those contracts were designed to accelerate the development and deployment of frontier AI for national security purposes, and they gave Google, operating through parent company Alphabet, an established contractual pathway into Pentagon networks.
The classified Gemini talks mark a new phase. Moving from unclassified administrative support to top-secret systems requires clearing stringent security accreditation processes that can take months or longer. Neither Google nor the Defense Department has publicly confirmed a timeline for when Gemini might be approved for classified use.
A crowded and contentious vendor field
Google is not the only AI company the Pentagon wants on its classified networks. Defense Secretary Pete Hegseth has publicly stated that the department intends to run multiple AI models on both unclassified and classified systems. In public remarks covered by the Associated Press, Hegseth said that Elon Musk’s Grok chatbot would operate alongside Google’s system inside Pentagon networks, covering both security tiers. The AP reported his comments as on-the-record statements rather than leaked internal directives. The strategy reflects a deliberate choice to avoid dependence on a single vendor.
But the multi-vendor approach has already generated friction. Hegseth warned Anthropic to let the military use the company’s AI technology on the Pentagon’s terms, according to AP sources familiar with the exchange. Anthropic CEO Dario Amodei pushed back publicly, saying his company “cannot in good conscience accede” to the Defense Department’s demands regarding contract language and the permissions it sought. That refusal drew a sharp line: at least one major AI vendor has decided the Pentagon’s conditions go too far.
Google has not made any comparable public statement. The company has not said whether it shares Anthropic’s concerns about safeguard removal, whether it is negotiating privately over guardrails, or whether it has accepted the Pentagon’s conditions without objection. That silence is itself notable. It suggests Google is either comfortable with the terms, working through disagreements behind closed doors, or simply unwilling to risk a public dispute that could jeopardize a high-value contract.
The Project Maven shadow
For Google, deeper military work carries institutional baggage. In 2018, the company faced an internal revolt when thousands of employees protested its involvement in Project Maven, a Pentagon program that used AI to analyze drone surveillance footage. Google ultimately chose not to renew the Maven contract and published a set of AI principles that included a pledge not to develop AI for weapons or surveillance that violated “internationally accepted norms.”
The company’s posture has shifted considerably since then. Google has expanded its defense and intelligence contracting, and its leadership has signaled greater willingness to work with the military. The classified Gemini talks represent the furthest extension of that shift, potentially placing Google’s most advanced AI inside the very systems its employees once objected to supporting.
What remains unresolved
Key details about the negotiations have not been made public. No official Google statement has addressed the scope of the classified deployment, and no primary Defense Department documents outlining specific use cases for Gemini on top-secret networks have surfaced. Whether the classified work would fall under the existing $200 million CDAO contract or require a separate agreement is unclear. That distinction matters: new contract vehicles typically trigger additional congressional notification and oversight requirements.
Unanswered safeguard questions as Gemini nears classified networks
The broader question hanging over the entire effort is what safeguards will govern commercial AI on classified military systems. The Pentagon wants maximum operational flexibility. Anthropic has publicly refused to grant it. Google, so far, has said nothing. Until more concrete documentation or on-the-record statements emerge, the guardrails that will or will not accompany Gemini’s entry into the Pentagon’s most sensitive networks remain an open question with significant national security implications.
More from Morning Overview
*This article was researched with the help of AI, with human editors creating the final content.
