The Pentagon gives 3 million staff their own versions of ChatGPT and Grok

Key points

• The U.S. Department of Defense has added ChatGPT Mil and Grok for Government to its secure AI portal, GenAI.mil, giving roughly 3 million civilian and military personnel access.
• GenAI.mil has already signed up more than 1.7 million unique users since launching last year with Google Gemini.
• Anthropic's Claude is absent after the Trump administration labeled Anthropic a supply-chain risk over a dispute about safety guardrails.
• Grok for Government reaches military users through SpaceX's Starshield AI, a secure satellite network built on Starlink technology.
• The Pentagon has also struck separate AI deals with Amazon Web Services, Microsoft, Nvidia and Reflection AI.

The U.S. Military now has its own version of two widely used AI chatbots. ChatGPT Mil and Grok for Government, custom builds of tools millions of civilians already use, went live this week on GenAI.mil, the Defense Department's secure AI access point.

What exactly is GenAI.mil?

It's a locked-down version of the public chatbot experience, one where government data doesn't flow through ordinary consumer channels. Standard consumer apps routinely collect usage data; the military portal is specifically exempt from that. The platform launched last year with Google Gemini and now carries three major AI systems.

As first reported by TechCrunch AI, more than 1.7 million of the department's roughly 3 million personnel have already created accounts. That's a faster adoption rate than most internal government software rollouts achieve in years.

What will troops and civilian staff actually use these tools for?

The two products have noticeably different pitches. ChatGPT Mil, from OpenAI's government programme, targets desk work: drafting documents, administrative tasks, logistics and planning on unclassified material.

Grok for Government, delivered through SpaceX's Starshield AI secure satellite network, is framed in broader operational terms. The Defense Department's press release says it will give the warfighter faster mission execution and stronger knowledge continuity across contexts from supply chain management to procurement research.

Why is Anthropic's Claude missing?

Claud's absence is pointed. Anthropic insisted on keeping certain safety guardrails in place before allowing unrestricted military access to its tools. The Trump administration responded by designating the company a supply-chain risk, a label that froze it out of this procurement. As we reported on 28 August, a federal judge has since ruled that blacklisting unconstitutional, a decision the Pentagon is still contesting.

This is one of the first public clashes over whether an AI company can set hard limits on how a government customer deploys its technology. It won't be the last.

What should ordinary people watch for?

If you work for the Defense Department, you now have access to two new AI tools through official channels. Use the portal, not a personal consumer account, because the official version keeps work data out of commercial pipelines.

For everyone else: the argument about safety guardrails versus unrestricted government access won't stay inside a single procurement fight. And it's worth remembering that Grok, before this deployment, had its own reliability problems. We covered a vulnerability that could expose user conversations, reported on 20 August, and a separate bug producing incoherent output the same week. Handing a tool with that recent a track record to 3 million military users is the detail this story probably undersells.