Skip to content

AI

Pentagon hands 1.7 million staff ChatGPT and Grok, minus the data grab

GenAI.mil now offers ChatGPT Mil and Grok for Government next to Gemini. The Pentagon says more than 1.7 million of its 3 million personnel are already on board. Claude is not on the guest list.

Cover photo: Jernej Furman from Slovenia / CC BY 2.0

TL;DR

  1. ChatGPT Mil and Grok for Government just joined Google Gemini on GenAI.mil.
  2. More than 1.7 million of 3 million DoD personnel are onboarded, per the department.
  3. Military versions skip consumer-style data collection; Anthropic's Claude is conspicuously absent.

What happened

The Department of Defense has added two custom tools to GenAI.mil: ChatGPT Mil and Grok for Government. They are tailored versions of the popular consumer products, now sitting inside a centralized, secure portal that launched last year.

GenAI.mil originally shipped with Google Gemini. The new arrivals widen the menu to three frontier models from three different companies.

The pitch is simple. Give Department of Defense employees access to commercial frontier AI without sending sensitive government data through ordinary consumer channels. The official military versions are also exempt from the data collection that is hard to dodge in consumer tech products.

That is the department's own figure, so treat it as a proud press-release number. Still, more than half the workforce has an account on a portal that is not even a year old.

The additions are part of what the Pentagon describes as a broader push to use AI to speed up work and keep an edge without sacrificing security.

There is also a subplot. The rollout highlights the absence of Anthropic's Claude, and the Pentagon's effort to work with other companies following its dispute with the lab. The Trump administration labeled Anthropic a supply-chain risk.

Why it matters

This is institutional adoption at scale, not a pilot with twelve volunteers and a slide deck. Millions of people who handle sensitive material now have sanctioned access to chatbots, rather than quietly pasting things into a consumer app and hoping for the best.

That is the real point of the portal. Staff were going to use these tools anyway. The question was whether their data would leave the building.

It also shows how vendor politics shape the market. Three models are in. One notable one is out, and the reason is a government dispute, not a benchmark.

For the AI labs, a captive audience of this size is a very good customer reference. For everyone else, it is a reminder that "frontier model" and "procurement" now live in the same sentence.

One caution: what we know comes from the department's own account. Usage figures are self-reported, and details on how the custom versions differ technically were not spelled out.

What to do with it

  • Check your own data path.If your team pastes work into consumer chatbots, find out whether an enterprise or restricted version exists.
  • Ask about data collection.The Pentagon's selling point is exemption from consumer-style data harvesting, so put that exact question to your vendor.
  • Don't bet on one model.GenAI.mil runs three vendors side by side, and your stack can too.
  • Watch the politics.Government disputes can remove a model from the menu overnight, so keep a fallback.

Ciao, sapiens. The military has a chatbot portal now. Your inbox still has no excuse.