> ## Content Index > Fetch the complete content index at: https://www.battlepolicy.com/llms.txt > Use this file to discover other available public pages before exploring further. # The Pentagon Put Three AI Vendors on One Platform. Their Terms Are Not the Same. - URL: https://www.battlepolicy.com/the-pentagon-put-three-ai-vendors-on-one-platform-their-terms-are-not-the-same/ - Published: 2026-09-01T05:51:19.000Z - Updated: 2026-09-01T05:51:19.000Z - Description: The Department of War put OpenAI's ChatGPT Mil and Starshield AI's Grok on GenAI.mil the same day, and the limits each vendor accepts now differ more than the models do. - Author: Marcus Schuler - Tags: USA, Autonomy, #daily-brief, OpenAI, xAI, Starshield AI, Anthropic, GenAI.mil *The Department of War put OpenAI's ChatGPT Mil and Starshield AI's Grok on GenAI.mil the same day, and the limits each vendor accepts now differ more than the models do.* The [Department of War](https://www.unite.ai/department-of-war-launches-openais-chatgpt-mil-on-genai-mil/?ref=battlepolicy.com) launched OpenAI's ChatGPT Mil on GenAI.mil on Aug. 31\. The custom system is accredited for Controlled Unclassified Information at Impact Level 5 and built to support more than 3 million personnel. It provides chat, file analysis, projects and custom GPTs, with further features sequenced over time. Frequently Asked Questions What is GenAI.mil? It is the Department of War's generative AI platform for unclassified work, launched Dec. 9, 2025 with Google Cloud's Gemini for Government. It has registered more than 1.7 million users from a workforce above 3 million, per the Department. What does Impact Level 5 accreditation mean? It certifies a system to handle Controlled Unclassified Information. Both ChatGPT Mil and Grok for Government carry it. Separate Department agreements cover classified networks at Impact Levels 6 and 7. What limits did OpenAI place on its military work? OpenAI published three red lines on Feb. 28: no mass domestic surveillance, no direction of autonomous weapons systems and no high-stakes automated decisions. It also said the deployment is cloud-only and that it retains discretion over its safety stack. Why was Anthropic designated a supply-chain risk? Anthropic insisted that mass surveillance of Americans and fully autonomous weapons stay off-limits, while the Department sought to use Claude for all lawful purposes. The Pentagon then designated the company a national security and supply-chain risk, per Axios. What did the court rule? U.S. District Judge Rita Lin vacated the designation on Aug. 27, writing in a 59-page decision that the measures imposed on Anthropic were "illegal and baseless." The government is expected to appeal. Which AI companies work on the Department's classified networks? The Department has agreements with eight frontier AI and cloud companies for Impact Levels 6 and 7: SpaceX, OpenAI, Google, NVIDIA, Reflection, Microsoft, Amazon Web Services and Oracle. AI-generated summary, reviewed by an editor. [More on our AI guidelines](https://www.battlepolicy.com/ai-guidelines/). Below this point 5 more sections. The [Department](https://ng.investing.com/news/stock-market-news/department-of-war-adds-starshield-ais-grok-to-genai-platform-93CH-2680420?ref=battlepolicy.com) added Starshield AI's Grok for Government the same day, also at Impact Level 5\. Grok offers Auto, Fast and Expert reasoning modes, persistent projects and reusable playbooks. Intended uses run from acquisition market research to logistics supply-chain management. GenAI.mil launched Dec. 9, 2025 with Google Cloud's Gemini for Government. Nine months on, it has registered more than 1.7 million users from a workforce above 3 million. The two additions make it a three-model platform. The models arrived under different limits. OpenAI published three red lines on Feb. 28: no mass domestic surveillance, no direction of autonomous weapons systems, no high-stakes automated decisions. The company kept the deployment cloud-only, retained discretion over its own safety stack and kept cleared personnel in the loop, while presenting its models as available for all lawful uses. A March 2 update added contract language barring surveillance of U.S. persons, and the Department affirmed that agencies such as the NSA would need a new agreement. Anthropic pressed for the same substance as binding contract terms, ruling out mass surveillance of Americans and fully autonomous weapons. The Department sought Claude for all lawful purposes and designated the company a national security and supply-chain risk. U.S. District Judge Rita Lin [vacated that designation](https://www.axios.com/2026/08/28/judge-blocks-pentagon-anthropic-blacklist?ref=battlepolicy.com) on Aug. 27, calling the measures "illegal and baseless" in a 59-page decision and noting the Pentagon kept pursuing work with Anthropic after imposing them. The substance of the limits overlaps. The instrument does not: OpenAI holds its safeguards as vendor policy, Anthropic wanted them written into the contract. The government is expected to appeal the Lin ruling, and Anthropic is fighting a second Pentagon designation in the D.C. Circuit.