The Pentagon has bolted two familiar consumer chatbots into its secure GenAI.mil portal, making tailored versions of ChatGPT and Grok available to the department’s workforce 3 million personnel.
GenAI.mil, the centralised portal that first offered Google’s Gemini when it launched last year, is intended to let Defence staff use frontier commercial models without routing sensitive work through ordinary consumer channels GenAI.mil. The department says the official military variants are not subject to the data collection typical of consumer products, a key reason for moving them behind government infrastructure exempt from data collection.
OpenAI’s build, named ChatGPT Mil, is presented as a familiar assistant: chat plus support for files, projects and custom GPTs, aimed at routine, document-heavy unclassified work such as administration, logistics, planning and policy chat, files, projects document-heavy routine work.
The Grok rollout comes through SpaceX’s Starshield AI and is pitched in broader operational terms. The Defence Department said, “Starshield AI’s Grok for Government will provide the warfighter with immediate productivity gains, stronger knowledge continuity and more secure and efficient collaboration.” The department also positioned Grok for tasks ranging from acquisition analysis to supply-chain management.
The move is an expansion rather than a new experiment: GenAI.mil has already attracted substantial use, with the department reporting more than 1.7 million unique users onboarded so far. The rollout fits a wider procurement push in which the Pentagon has struck deals with several commercial cloud and AI suppliers, including Amazon Web Services, Microsoft and Nvidia.
Notably absent from the new mix is Anthropic’s Claude; the department’s engagement with front-line labs has been shaped by a public dispute over usage and safety guardrails, after Anthropic was branded a supply-chain risk by the prior administration and is contesting that designation in court labeled a supply-chain risk. That disagreement has influenced which vendors appear in the secure environment and under what terms they will operate.
For Defence employees, the change is practical: generative AI will now be available for day‑to‑day unclassified tasks inside a government-controlled service rather than via consumer apps. For external observers and the workforce alike, the next questions are which features the services add over time and how the department manages vendor risk and model behaviour as use spreads.
