Before ChatGPT was ever available on it, the Pentagon’s AI portal had already seen roughly 1.7 million logins. Set against the Department of Defense’s 3 million civilian and uniformed personnel, that is over half the workforce using a product that barely registered as existing a year ago.
The department has now bolted ChatGPT Mil and Grok for Government onto GenAI.mil, the centralized secure portal it stood up last year. Each is a bespoke version of the consumer product, tailored to what officials describe as “warfighter needs.”
The absence is louder than the announcement
Claude is not on the list. Anthropic’s model is conspicuously missing from a portal that now hosts offerings from both OpenAI and xAI — and that gap is deliberate, not clerical.
The Trump administration branded Anthropic a supply-chain risk after the company declined to hand the Pentagon unrestricted access to its AI tools, insisting on certain safety guardrails instead. Anthropic is contesting the designation in court.
The new additions therefore serve two purposes at once: they push the department’s AI agenda forward, and they signal that the Pentagon is not short of alternatives following its falling-out with the frontier lab.
What GenAI.mil is supposed to solve
Google Gemini was the portal’s launch model. The reasoning behind it was simple enough — put commercial frontier models in the hands of Defense Department staff without pushing sensitive government material through the usual consumer pipes.
Unlike consumer tech products, where data collection is difficult to avoid, the official military build is exempt from it. That exemption is arguably the whole value proposition. Strip it away and you are left with a model anyone can already run at home.
ChatGPT Mil is the boring one, and that’s fine
ChatGPT Mil arrives via OpenAI for Government, and anybody familiar with the commercial product will recognize it immediately. The department lists chat, files, projects and custom GPTs.
The advertised use cases skew toward routine, unclassified, document-heavy labor: administrative tasks, logistics, planning and policy. Additional features are slated to arrive later.
Nothing is promised about missions or precision here. It is paperwork — which, for most of those 3 million people, is what the workday actually consists of.
Grok got the war-movie treatment
xAI’s tool drew markedly different rhetoric from the Pentagon.
“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’s press release said. Starshield AI refers to SpaceX's secure satellite network, which is built on the company’s existing Starlink technology.
Grok, the department went on, would enable its military “to execute missions faster and with greater precision across numerous operational contexts, ranging from market research analysis for acquisition professionals to supply-chain management for logisticians.”
Go back over those examples. The operational contexts on offer are market research and supply-chain management. Battlefield language wrapped around procurement work.
The vendor list keeps growing
Amazon Web Services, Microsoft, Nvidia and Reflection AI have all signed agreements with the Pentagon under the same initiative.
For someone working inside the department, the useful question is not which model has the highest IQ. It is whether the secure ChatGPT sitting on the work machine handles the task you would otherwise paste into the consumer app on your phone — because stopping exactly that habit is the reason GenAI.mil exists.