The Pentagon opened OpenAI's ChatGPT Mil and Starshield AI's Grok for Government to 3 million military and civilian personnel on Monday, leaving Anthropic's Claude off the platform.
Key Points:
- ChatGPT Mil and Grok for Government joined Google's Gemini on GenAI.mil, the Pentagon's internal portal for unclassified generative AI work.
- Both models cleared Impact Level 5, the department's highest authorization tier for sensitive but unclassified data.
- Claude remains absent, days after a federal judge struck down the Pentagon's supply-chain risk label as unlawful retaliation.
GenAI.mil Adds ChatGPT Mil And Grok
Defense Department officials confirmed the rollout Monday, saying staff can now pick among three commercial frontier models for unclassified work once they log into GenAI.mil, the department's internal generative AI portal.
The portal opened in December with Google's Gemini for Government as its only option, and it drew relatively few users during its first months. More than 1.7 million people have signed on since.
Both new models cleared Impact Level 5, the department's highest authorization tier for storing and processing sensitive information that has not been classified, according to the two press releases issued Monday.
That tier dictates where the data sits and who may handle it.
ChatGPT Mil centers on chat, files, projects and custom GPTs, and the department said it will carry document-heavy work across planning, policy, logistics and administration. Grok for Government adds adaptive reasoning modes, persistent workspaces and reusable playbooks that are meant to spread institutional knowledge across scattered units and commands. Officials described that tool in far more operational language.
Also Read: Kalshi Bans George Santos For Life With A $71,356 Fine Over State Of The Union Bets
Warfighter Framing And Unanswered Limits
Joe Larson, OpenAI's vice president for government and a former senior Pentagon official, framed the arrangement as pairing capable tools with clear safeguards and human judgment on consequential decisions. A War Department official said the department wants an architecture that keeps it from locking itself into any single commercial AI vendor over the long term.
The limits stayed vague. A Pentagon official declined to say whether ChatGPT Mil can support targeting, intelligence analysis or decisions that involve the use of force, pointing instead to the certification. The department has signed separate agreements with Amazon Web Services, Microsoft, Nvidia and Reflection AI to push similar tools onto classified military networks.
Anthropic Dispute Behind Claude's Absence
Claude was slated to join the same portal until Anthropic refused to lift contract terms barring mass surveillance of Americans and the use of its models in lethal autonomous weapons. The standoff never eased.
Defense Secretary Pete Hegseth designated the company a supply-chain risk in February, and Donald Trump ordered agencies well beyond the military to stop using Claude altogether.
Anthropic sued and won a reversal last week. U.S. District Judge Rita Lin ruled on Aug. 27 that the label amounted to unlawful First Amendment retaliation, ordered it removed and faulted the government's stated reasoning, while a second case stays pending in Washington.
Read Next: Hyperliquid Laundering Risk Uncovered In $30M Lazarus Trades
