Pentagon adds Google's Gemini 3.1
- The Pentagon added Google’s Gemini for Government 3.1 Pro and 3.0 Flash to GenAI.mil on April 27 after a preview rollout. - Defense officials said the new models cut hallucinations, improved output quality, and are now available to more than 3 million users. - The move lands as 600-plus Google employees oppose classified Pentagon AI work. (defenseone.com)
The Pentagon has added Google’s Gemini for Government 3.1 Pro and 3.0 Flash models to GenAI.mil, its internal generative artificial intelligence platform for military and civilian users. (defenseone.com) (placera.se) The Office of the Under Secretary of War for Research and Engineering announced the change on April 27, saying the upgrade would reduce hallucinations, improve output quality, and speed up work for warfighters. (placera.se) (tipranks.com) Defense One reported that Gemini 3.1 Pro had already been available in preview for several weeks before the Pentagon made the addition official. The model is also set to be available to Gemini for Government users across the federal government. (defenseone.com) GenAI.mil launched on December 9, 2025, with Google Cloud’s Gemini for Government as its first frontier model. The Pentagon said at launch that the system would reach civilians, contractors, and military personnel on desktops across the Pentagon and U.S. bases worldwide. (war.gov) The platform now serves more than 3 million users, according to reports on this week’s rollout. That scale turns Google’s model update into a department-wide expansion of commercial artificial intelligence inside the Defense Department. (stocktwits.com) (defenseone.com) At the same time, more than 600 Google employees signed a letter to Chief Executive Sundar Pichai urging him to reject classified military AI work with the Pentagon. The letter said the company could not guarantee its systems would avoid “unmonitored harm.” (thehill.com) (mercurynews.com) The Hill reported that the employee letter followed talks over a separate classified agreement that would let the Pentagon use Google’s AI for “all lawful purposes.” Reuters, citing The Information on April 28, reported that Google has now joined other firms supplying models for classified Pentagon work. (thehill.com) (usnews.com) Google’s government product has also been tailored for federal compliance. Google Cloud documentation says Gemini for Government can be deployed to meet FedRAMP High and Defense Department Impact Level 4 requirements. (docs.cloud.google.com) Google introduced Gemini 3.1 Pro on Google Cloud on February 19 and described it as its most advanced model for complex tasks. Its model card says it is designed for large, multimodal reasoning problems that combine text, images, audio and video. (cloud.google.com) (deepmind.google) The Pentagon’s latest move extends a relationship that began with unclassified work on GenAI.mil and now sits alongside reported classified-use agreements with other leading AI labs. The fight inside Google shows that the commercial race to sell frontier models to defense agencies is still colliding with employee limits on how those tools should be used. (war.gov) (usnews.com)