Overview
- The Department of Defense put tailored versions of OpenAI’s ChatGPT and xAI/Starshield’s Grok on GenAI.mil Monday, making the tools available to military and civilian personnel for controlled unclassified information work.
- Both systems carry Impact Level 5 accreditation, which permits handling Controlled Unclassified Information and lets users run planning, policy, logistics, administration and other document-heavy tasks inside a hardened government cloud.
- The department said more than 1.7 million unique users have already accessed GenAI.mil since its December launch and that offering multiple models is meant to avoid vendor lock-in and give users model choice.
- OpenAI and the Pentagon say data processed on GenAI.mil will remain isolated from commercial model training, but Anthropic’s Claude remains absent after a public dispute and a federal judge recently ruled the government’s supply-chain designation of Anthropic unlawful.
- GenAI.mil was built partly to stop unauthorized 'shadow AI' use and centralize safe tools for routine work while the Pentagon continues separate efforts to develop classified AI systems for combat and warfighting.