Pentagon Adds ChatGPT and Grok to Its AI Toolkit — and Still Won't Touch Claude

The Pentagon just added custom versions of OpenAI's ChatGPT and xAI's Grok to GenAI.mil, its genAI military platform for roughly 3 million civilian and military personnel. ChatGPT Mil and Grok for Government join Google's Gemini as the third and fourth frontier models on the department's secure AI portal. The department says 1.7 million of its 3 million personnel have already signed on. One name is conspicuously missing from that list: Anthropic's Claude, still frozen out after a public standoff over how the military is allowed to use it.

What just launched: ChatGPT Mil and Grok for Government

ChatGPT Mil and Grok for Government are custom, government-hosted versions of the consumer tools. Both are built to keep sensitive Defense Department data out of ordinary commercial channels. ChatGPT Mil came out of OpenAI's OpenAI for Government program. According to TechCrunch's reporting on the Defense Department's rollout, it focuses on chat, files, projects, and custom GPTs, aimed at document-heavy, unclassified work like administrative tasks, logistics, planning, and policy. Grok for Government is pitched in more overtly military terms. It's accredited for Controlled Unclassified Information at Impact Level 5, and the Department of War says it brings deep-thinking inference, adaptive reasoning modes, persistent projects, and reusable "playbooks" that capture institutional knowledge.

"Starshield AI's Grok for Government will provide the warfighter with immediate productivity gains, stronger knowledge continuity and more secure and efficient collaboration." Department of War press release

The department frames Grok's use cases broadly, from market research analysis for acquisition staff to supply chain management for logisticians, rather than tying it to any single job function.

What is GenAI.mil, and who's actually using it?

GenAI.mil is the department's answer to a real problem. Staff were already using commercial AI tools on personal devices, and the Pentagon wanted that traffic on government-controlled infrastructure instead. It launched with Gemini for Government as the first model on December 9, 2025, certified for Controlled Unclassified Information at Impact Level 5.

"The future of American warfare is here, and it's spelled AI." Pete Hegseth, Secretary of War

Adoption has scaled quickly since. By May 2026, the platform reported 1.3 million users. By the ChatGPT and Grok rollout in late August, that had grown to over 1.7 million. Usage isn't just chat, either. According to Defense One's reporting, GenAI.mil users have already built more than 100,000 AI agents through Agent Designer, with authorization to work on the department's most sensitive unclassified data.

Why isn't Claude on the list?

Anthropic's absence traces back to a specific policy fight, not an oversight. CEO Dario Amodei said Anthropic would not drop two safeguards from its Pentagon contracts: mass domestic surveillance and fully autonomous weapons that take humans out of the loop entirely. Secretary Hegseth argued the department shouldn't be bound by any vendor's usage policy.

"Here's what we're asking: Allow the Pentagon to use Anthropic's model for all lawful purposes... We will not let ANY company dictate the terms regarding how we make operational decisions." Sean Parnell, Pentagon chief spokesperson

When Anthropic held its ground, the Trump administration directed federal agencies to stop using its technology and designated the company a supply chain risk. A district court granted Anthropic a preliminary injunction in March, but an appeals court denied a stay and undid that injunction in April, leaving the legal fight unresolved. That tangle sits alongside the earlier order to cut off Claude Fable 5 and Mythos, part of a broader pattern of friction between Anthropic and federal agencies.

The Pentagon's bigger AI buildout

GenAI.mil is only the visible layer of a much larger push. In May 2026 the department signed agreements with eight AI companies, including SpaceX, OpenAI, Google, Nvidia, Reflection, Microsoft, AWS, and Oracle, to deploy their tools on classified Impact Level 6 and 7 networks. The stated goal is to avoid vendor lock-in by keeping multiple frontier AI providers in play at once.

"We believe the people responsible for defending the country should have access to the best tools available." OpenAI, Bringing ChatGPT to GenAI.mil

That diversification strategy is also a hedge. Sachin Seth, a VC at Trousdale Ventures who focuses on defense tech, said losing Anthropic entirely would leave a real gap.

"They would have to wait six to 12 months for either OpenAI or xAI to catch up. That leaves a window of up to a year where they might be working from not the best model, but the second or third best." Sachin Seth, Trousdale Ventures

That gap is part of why OpenAI and xAI moved quickly to fill the space Anthropic left open.

The pushback: surveillance and autonomous weapons concerns

Not everyone is convinced the guardrails on these new deployments are strong enough. The Electronic Frontier Foundation has argued OpenAI's own commitments don't go far enough, calling the safety language "weasel words" that won't meaningfully stop AI-powered surveillance use cases. The underlying legal gap is real. There's currently no binding U.S. law that categorically prohibits AI-directed autonomous weapons or surveillance, so restrictions exist only where an individual vendor's contract writes them in.

That's precisely the tension Anthropic's fight highlighted. It's why the debate over GenAI.mil's new additions isn't really about chat interfaces. It's about which company's usage policy, if any, actually governs what the Pentagon can do with the AI running on its network.