The Pentagon just got ChatGPT and Grok. The press notes keep saying warfighter. The accreditation is IL5 unclassified.
Today — Monday, 31 August 2026 — the US Defense Department launched OpenAI’s ChatGPT Mil and Starshield AI’s Grok for Government on GenAI.mil, joining Google Gemini on a portal built for three million civilian and military personnel. GenAI.mil itself went live on 9 December 2025 with Gemini for Government. That’s the baseline. Today is the expansion.
I don’t need a victory lap. I need the boundary.
Workplace AI. Not combat.
Military Times, citing Defense One, is blunt: Pentagon and OpenAI officials do not intend to use the models on GenAI.mil for combat. Classified and warfighting AI sit somewhere else. This stack is accredited at Impact Level 5 for Controlled Unclassified Information — CUI. Sensitive. Still unclassified. Chat, files, planning memos. Not a targeting loop.
DoD says the portal has already onboarded more than 1.7 million unique users. That’s from the Grok announcement, via MeriTalk and Military Times. A lot of people. Still not a weapons system.
From Hermosillo that reads as: the Pentagon bought a very expensive office assistant and dressed it in warfighter copy. The copy isn’t fake. The product just isn’t what the word sounds like.
ChatGPT Mil is the familiar one
ChatGPT Mil comes out of OpenAI for Government. Defense One quotes the department: it “brings a familiar commercial experience into the Department’s secure environment, tailored to warfighter needs.” Core features are the ones you already know — chat, files, projects, custom GPTs — with more coming later. The job, in DoD’s words, “supports document-heavy unclassified work across the Department, including planning, policy, logistics, and administration, and is built for scale to support more than 3 million Department personnel. This deployment accelerates routine tasks and frees up time for more critical projects across the Joint Force.”
MeriTalk has the ecosystem line: “Integrating ChatGPT Mil into GenAI.mil alongside existing frontier AI capabilities establishes a robust, multi-model ecosystem for the warfighter.”
Joe Larson, OpenAI’s VP and head of government, is on Defense One with the safeguards pitch: “This integration is grounded in our national security principles, pairing access to powerful tools with clear safeguards, meaningful human judgment, and accountability for high-stakes decisions.” And: “ChatGPT on GenAI.mil will make unclassified work throughout the Department more efficient, freeing up time and resources so DoD personnel can focus on the mission… We’re excited to see what the Department’s extraordinary workforce builds with it.”
OpenAI already said on 9 February 2026 it would bring a custom ChatGPT to GenAI.mil for unclassified work, with data isolated from commercial training. That was a next-phase announcement. Today is the go-live — ChatGPT Mil and Grok, together.
Grok, vendor lock, and the missing lab
DefenseScoop and MeriTalk put Grok for Government as an xAI Grok frontier model tailored for federal use, shipped as Starshield AI’s product. TechCrunch’s gloss: SpaceX’s Starshield AI is a secure satellite network that uses existing Starlink tech. I’m not here to unpack the corporate chart. The Pentagon’s line is about the desk.
DoD, via TechCrunch and MeriTalk: “Starshield AI’s Grok for Government will provide the warfighter with immediate productivity gains, stronger knowledge continuity and more secure and efficient collaboration.” It also says Grok lets the Joint Force “execute missions faster and with greater precision across numerous operational contexts, ranging from market research analysis for acquisition professionals to supply-chain management for logisticians.” Read that list again. Market research. Supply chain. Logisticians. That’s the product.
Capabilities, per DoD (DefenseScoop / MeriTalk): deep-thinking inference; adaptive reasoning modes Auto, Fast, and Expert; customizable workspaces; persistent projects; reusable “playbooks” that capture and scale institutional knowledge. The expansion, DoD says via MeriTalk, “provides another top-tier generative AI tool for warfighter advantage, eliminating vendor lock and promoting a vibrant American AI ecosystem.”
Claude isn’t on the list. TechCrunch notes Anthropic was labeled a supply-chain risk after it refused unrestricted Pentagon use and insisted on safety guardrails; it’s fighting that designation in court. DefenseScoop reported that last week a federal district judge ruled the Pentagon’s designation and actions against Anthropic were “illegal and baseless.” That’s the aside. It’s not today’s launch.
On data: TechCrunch says the official military version is exempt from the collection that’s hard to dodge in consumer products. Military Times reports OpenAI’s line that data processed on GenAI.mil stays isolated in the government environment and is not used to train or improve public or commercial models.
I’m not scoring the branding. I’m noting the split. GenAI.mil is now a multi-model workplace — Gemini, ChatGPT Mil, Grok — for unclassified CUI at IL5. The Pentagon can call it warfighter advantage. The paperwork still says planning, policy, logistics, and administration. Those are different sentences. Both got published today.
