The Department of War announced on August 31 that ChatGPT Mil and Grok for Government are live on GenAI.mil, the Pentagon's enterprise generative AI platform, per Military Times. Both hold accreditation for Controlled Unclassified Information at Impact Level 5. Grok arrives through Starshield AI and reaches more than 1.7 million personnel out of a workforce near 3 million, per NOTUS. They join Google's Gemini, which has been on the platform since its December launch, per DefenseScoop.
The accreditation is the part that carries weight. Impact Level 5 covers Controlled Unclassified Information whose unauthorized disclosure could cause serious harm to national security, which is a materially higher bar than the pilots most agencies have run. Reaching it means the models are cleared for real work rather than sandboxed experiments, and the 1.7 million user figure indicates deployment at genuine enterprise scale rather than a limited trial.
The procurement design is deliberate and worth reading closely. Putting three competing frontier models on one platform is an explicit move against vendor lock-in, and it follows a pattern we have tracked all year: Apple routing Siri through Gemini, Microsoft building the MAI family to reduce dependence on OpenAI, enterprises architecting for model portability after Commerce forced Anthropic's models offline in the spring. The largest single buyer of software in the world has now written that same hedge into its reference architecture.
One vendor is conspicuously missing, and the reason is a live legal fight rather than a technical gap. The Department designated Anthropic a supply-chain risk after the company insisted on contractual guardrails barring its models from mass surveillance of Americans and from lethal autonomous weapons, a designation that effectively cut it off from federal contracts. A federal judge struck that designation down last week, ruling the government had unlawfully retaliated. So the Department is assembling a multi-vendor stack at precisely the moment the courts are defining how much say it gets over each vendor's safety policies, and the one company that drew a hard line is currently litigating its way back onto the platform.
Bottom Line
IL5 accreditation and 1.7 million users make this a production deployment rather than a pilot. Watch the Anthropic litigation, because it is setting how far the government can go in penalizing a vendor for the safety limits it writes into its own contracts.