The Pentagon’s GenAI.mil platform just got a lot more interesting. After launching last December with only Google Gemini as its flagship model, the Department of Defense is now adding OpenAI’s ChatGPT Mil and xAI’s Grok for Government to the mix. The platform has already attracted more than 1.7 million users across the DoD’s three-million-plus workforce, which tells you how quickly secure AI tools are becoming standard-issue infrastructure rather than experimental toys.
The two new additions come with clearly differentiated roles. ChatGPT Mil is positioned for the unglamorous but essential work: administrative tasks, logistics, planning. Grok, meanwhile, gets the more defense-flavored pitch, with the Pentagon highlighting procurement analysis and supply chain management as its sweet spots. That division of labor is telling. Rather than treating these as interchangeable chatbots, DoD is signaling that different models will earn their keep in different operational corners.
Security remains the non-negotiable. Both models run in a dedicated environment completely walled off from their commercial counterparts, meaning no sensitive government data flows back to OpenAI or xAI. User data isn’t collected, which addresses the most obvious objection to putting classified-adjacent workflows through third-party models.
Conspicuously absent is Anthropic’s Claude. The company refused to grant the Pentagon unrestricted use, and the Trump administration subsequently labeled it a supply chain risk. A federal court struck down that classification as unlawful in late August, but Claude still hasn’t made it onto the platform. Whether that changes under legal pressure or stays frozen by bureaucratic inertia is an open question. For now, the Pentagon’s AI portfolio is a three-model race, and the missing fourth player says as much about AI politics as it does about technical capability.
