In This Article
Key Takeaways
- On August 31, 2026, the Department of War launched OpenAI's ChatGPT Mil and Starshield AI's Grok for Government on GenAI.mil, alongside Google's Gemini for Government, which went live in February.
- All three models are accredited at Impact Level 5 for Controlled Unclassified Information, so they can handle sensitive unclassified work but nothing classified.
- GenAI.mil has drawn more than 1.7 million unique users out of a 3-million-plus workforce in about nine months, and users had already built over 100,000 custom agents by July.
- Anthropic's Claude is not on the platform following the department's supply-chain-risk designation of the company, a dispute that is still in litigation.
What launched on August 31
The Pentagon's enterprise generative AI portal just got two new models. On Monday, August 31, 2026, the Department of War switched on OpenAI's ChatGPT Mil on GenAI.mil, and in the same announcement added Starshield AI's Grok for Government, per DefenseScoop. They join Google Cloud's Gemini for Government, which DefenseScoop notes was the first model on the platform when it went live in February 2026.
ChatGPT Mil is a custom build of OpenAI's product for the department. TechCrunch reported that it ships with chat, file handling, projects, and custom GPTs, and that the department is pointing it at document-heavy, routine unclassified work such as administrative tasks, logistics, planning, and policy. Military Times listed disaster response, resource allocation, supply management, and transportation requirements as intended uses. DefenseScoop reported the deployed model is currently GPT-5.4 Terra, with GPT-5.6 Terra to follow.
Grok for Government is positioned more broadly. TechCrunch quoted the department's framing that it gives the warfighter immediate productivity gains, stronger knowledge continuity, and more secure and efficient collaboration, spanning market research, supply-chain management, and operational contexts.
IL5 and CUI, read plainly
The detail that matters most for anyone who works with or sells to the department is the accreditation level. Per Military Times, ChatGPT Mil is accredited for Controlled Unclassified Information at Impact Level 5, the DoD cloud-security standard for sensitive unclassified data. DefenseScoop says all three models on GenAI.mil hold IL5 clearance, which authorizes storage and processing of non-public, sensitive unclassified data.
That is a real ceiling. IL5 covers CUI and unclassified national security system data, and it requires physical separation from non-DoD, non-federal tenants, as Precision Federal's plain-language explainer of what IL4, IL5 and IL6 change for an AI workload lays out. Nothing classified goes into these tools today. DoD's chief digital and AI officer, Cameron Stanley, told an AWS Summit audience in July that the department is looking at GenAI.mil going to higher classification levels, per Nextgov, but that is a plan, not a delivered capability.
Military Times also reported the data-handling terms: user data stays isolated inside the government environment and is not used to train OpenAI's public or commercial models, and the deployment required security testing by the Defense Information Systems Agency and the NSA. OpenAI's Joe Larson, its vice president and head of government, told DefenseScoop the integration pairs access to powerful tools with clear safeguards, meaningful human judgment, and accountability for high-stakes decisions.
How big GenAI.mil already is
The user numbers are the reason this launch matters beyond the Pentagon. Military Times reported that GenAI.mil has attracted more than 1.7 million unique users in the roughly nine months since it launched, against a total workforce of more than 3 million military members, civilian employees, and contractors. Nextgov reported the same 1.7 million figure in July and added that users had already created more than 100,000 custom agents on the platform.
For context, that is well over half the eligible workforce using an enterprise AI tool inside a year, which is faster adoption than most Fortune 500 rollouts we have covered. It also means the model choice inside GenAI.mil now shapes daily habits for a very large share of federal knowledge workers. Nextgov listed SpaceX, OpenAI, Google, NVIDIA, Reflection, Microsoft, Oracle, and Amazon Web Services among the vendors whose capabilities are available on the platform.
Who is not on the platform
One frontier lab is conspicuously absent. Both TechCrunch and DefenseScoop reported that Anthropic's Claude is not on GenAI.mil after the department designated the company a supply-chain risk earlier this year. DefenseScoop's account is that officials had said in December 2025 that Claude would follow the other models, that in February 2026 Anthropic sought contractual limits on mass-surveillance and lethal-autonomous-weapons uses while the department wanted unrestricted access for lawful purposes, and that Anthropic subsequently sued. We are not summarizing the litigation here; the point for practitioners is simply that the current menu is Gemini, ChatGPT Mil, and Grok.
What to do if you work in or sell to DoD
Three practical reads. First, if you are a contractor whose work product will be consumed inside GenAI.mil, assume your documents will be summarized, queried, and turned into custom agents by the government side. Write for that: clear headings, explicit requirements tables, and no facts buried in prose.
Second, if you build AI for the department, the IL5 line is the design constraint. A tool that needs classified data to be useful cannot ride on GenAI.mil today. A tool that works on CUI can, and now has a 1.7-million-user distribution channel it did not have a year ago.
Third, watch the model list. The department added two models in one day and its CDAO has said more are coming. The portal, not any single vendor, is the durable asset, and the acquisition pattern here, an enterprise partnership first and a platform rollout months later, is the one to expect for the next additions. The official release describes the ChatGPT Mil launch as the formal adoption phase of an enterprise partnership established with OpenAI in 2025.
Sources: Military Times — The military's ChatGPT is now live via the Pentagon's GenAI platform; DefenseScoop — Grok and ChatGPT join Gemini in Pentagon's enterprise genAI portal; TechCrunch — The Pentagon now has its own version of ChatGPT and Grok; Nextgov/FCW — GenAI.mil records almost 1.7M users, plans new model additions; U.S. Department of War — Launches OpenAI's ChatGPT Mil on GenAI.mil. Analysis and framing by Precision AI Academy.
Common questions
Which AI models are on GenAI.mil now? As of August 31, 2026: Google's Gemini for Government, OpenAI's ChatGPT Mil, and Starshield AI's Grok for Government, per DefenseScoop and Military Times.
Can GenAI.mil handle classified data? No. The models are accredited at Impact Level 5 for Controlled Unclassified Information. DoD's CDAO has said the department is looking at higher classification levels, but that has not shipped.
Is user data used to train OpenAI's public models? Military Times reported that data stays isolated inside the government environment and is not used to train OpenAI's public or commercial models.
How many people use GenAI.mil? More than 1.7 million unique users out of a workforce of over 3 million, with more than 100,000 custom agents created, per Military Times and Nextgov.