ChatGPT Mil and Grok for Government Launch at the Pentagon

Today the Pentagon flipped on its biggest frontier-AI switch yet: ChatGPT Mil and Grok for Government are live on GenAI.mil, the DoD's secure portal. Roughly 3 million personnel are eligible — and over 1.7 million are already onboarded.

What shipped

ChatGPT Mil, from OpenAI for Government, covers chat, files, projects, and custom GPTs — aimed at unclassified, document-heavy work: administrative tasks, logistics, planning, policy. Grok for Government, delivered through SpaceX's Starshield AI, comes with heavier warfighter framing: "execute missions faster and with greater precision," from acquisition market research to supply-chain management. Both are custom builds that skip the data collection baked into consumer products. GenAI.mil launched last year with Google Gemini; it now runs the two biggest consumer models in uniform.

Why it matters

The absence is the headline: Anthropic isn't in the room. The administration labeled it a supply-chain risk after it refused unrestricted Pentagon use and insisted on safety guardrails — a designation it's fighting in court. Translation for anyone watching procurement: holding the line on AI safety is now a competitive liability, and the strictest lab just got iced out of a 3-million-seat deployment.

Then there's velocity: 1.7M unique users on GenAI.mil before these models even landed. DoD doesn't run pilots anymore. And connect the dots — SpaceX runs Grok for Government while closing its Cursor acquisition, the same week OpenAI cut Cursor's model access. The defense AI stack is consolidating around one orbit, and the guardrails-first lab is officially outside it.

Verdict: watch Anthropic's court fight. It's the referendum on whether safety posture survives contact with government procurement.