The Pentagon Just Gave 3 Million Troops ChatGPT and Grok. Claude Is the Only Major AI Missing — Here’s Why.

The US Department of Defense today announced it has added OpenAI’s ChatGPT and xAI’s Grok to GenAI.mil — a secure internal platform that gives military and civilian DoD personnel access to commercial AI tools on a non-classified network. With today’s expansion, the platform is now available to all 3 million DoD personnel, with 1.7 million unique users already onboarded.

It is one of the largest AI deployments in history — measured by the size of the workforce being handed AI access in a single institutional move. And the most politically charged detail is not who is on the platform. It’s who isn’t.

Anthropic’s Claude — the AI model that multiple analysts currently rank as the leading enterprise AI product — is notably absent. And the reason illuminates something important about where AI governance is heading.

What GenAI.mil Actually Is

GenAI.mil launched in December 2025 with a single model: Google’s Gemini. The platform is gated by DoD Common Access Cards and restricted to government-furnished equipment on the military’s unclassified network. It is approved to handle controlled unclassified information at Impact Level 5 — a standard that requires FedRAMP High certification, physical tenant separation, and US-person staffing restrictions. This is not a consumer chatbot deployment. It is a hardened government AI environment built to strict federal security standards.

The Air Force went a step further last year, decommissioning its own in-house chatbot, NIPRGPT, and instructing 700,000 users to migrate directly onto GenAI.mil. The centralisation of the DoD’s AI tooling onto a single managed platform is itself a significant institutional move — one that makes the choice of which models appear on that platform a high-stakes decision.

Today’s expansion adds two new models. ChatGPT Mil — developed through OpenAI’s government programme — handles “document-heavy unclassified work across the Department, including planning, policy, logistics, and administration.” Grok for Government — deployed through Starshield, SpaceX’s secure satellite network for government entities — is positioned for “immediate productivity gains, stronger knowledge continuity, and more secure and efficient collaboration,” according to the DoD press release.

Why Claude Isn’t There

Anthropic was part of the original GenAI.mil prototype awards in 2025. Claude was planned as a fourth model on the platform alongside Gemini, ChatGPT, and Grok. It never made it.

In February 2026, the Pentagon labelled Anthropic a “national security supply chain risk” — a statutory designation previously reserved for companies with ties to foreign adversaries like Huawei and ZTE. The reason: Anthropic refused to strip safety guardrails from Claude that prohibited the model from being used in fully autonomous weapons systems or for domestic mass surveillance. The company declined to provide the Pentagon with unrestricted access.

The designation had consequences. It blocked Anthropic from certain government contracts and created a cloud over the company’s ability to work with federal agencies. Anthropic fought back in court — and won. On August 27, just five days ago, Judge Rita Lin issued a permanent injunction barring enforcement of the supply chain risk designation, ruling it was unconstitutional retaliation for Anthropic’s public ethical positions.

But despite winning in court, Claude remains off GenAI.mil. Pentagon Chief Technology Officer Emil Michael confirmed this week that the department expects to complete the removal of Anthropic platforms by late September — suggesting that even after the court ruling, the DoD intends to proceed with exclusion rather than inclusion. A parallel case under a different federal statute is still pending in the DC Circuit.

The Grok Complication

There is a significant and underreported dimension to the Grok deployment that deserves attention. Before the DoD finalized the contract with xAI, the company’s own engineers reportedly told the Pentagon that Grok had no reliable technical fix for generating child sexual abuse material. The DoD proceeded with the deployment anyway.

Federal law under 18 U.S.C. § 2258A requires any electronic service provider that becomes aware of CSAM on its platform to report it to the National Center for Missing and Exploited Children’s CyberTipline as soon as reasonably possible. The Take It Down Act, in effect since May 2026, further authorises FTC fines of $53,088 per image not removed within 48 hours of a valid takedown request. Neither xAI nor the DoD has publicly addressed what reporting mechanism, if any, would apply if Grok for Government generates CSAM on a DoD-operated network.

This is not a small detail. It raises genuine questions about the diligence applied to deploying a model with a known, unresolved safety issue on a platform used by 3 million government employees — while simultaneously excluding a model whose only offence was insisting on safety guardrails.

What ChatGPT Mil and Grok Will Actually Be Used For

Setting aside the politics, the practical use case for GenAI.mil is mundane in the best possible way. The DoD employs millions of people who spend significant portions of their working hours on administrative tasks: writing reports, reviewing documents, analysing logistics data, drafting policy guidance, managing supply chains, and processing procurement paperwork.

Generative AI is genuinely useful for all of these tasks. The Pentagon’s stated goal is to free up time and cognitive bandwidth so that personnel can focus on higher-value, mission-critical work rather than document formatting and routine correspondence. For a workforce of 3 million, even modest productivity improvements aggregate into significant capability gains.

ChatGPT Mil offers features familiar to commercial users: chat, file handling, projects, and custom GPTs. Grok for Government is positioned for market research, acquisition analysis, and supply-chain management. The models operate only on unclassified data — they are not plugged into classified systems, operational intelligence feeds, or weapons targeting infrastructure.

The Broader Significance

The GenAI.mil expansion is more than a procurement story. It’s a signal about how the world’s most powerful military intends to integrate AI into its operations — and about the terms on which it expects AI companies to participate.

The implicit message from today’s announcement is clear: AI companies that accept the DoD’s terms get access to one of the most significant institutional deployments in the world. Companies that insist on safety restrictions — even ones a federal court has ruled are legally protected — find themselves excluded.

That tension is not resolved by today’s news. Anthropic’s court victory is real and meaningful. But Claude’s absence from GenAI.mil, even after the ruling, shows that legal victories don’t automatically translate into restored relationships. The policy fight over what AI companies owe their government customers — and what they can refuse — is very much ongoing. For more on Anthropic’s legal battle and its IPO ahead, see our coverage of Anthropic’s record Q2 revenue and the OpenAI frontier training pause that reshaped the AI safety debate this month.

Read more tech related articles here.