The U.S. Department of War added government-tailored versions of ChatGPT and Grok to its GenAI.mil platform this week, giving its 3 million military and civilian employees access to tools beyond the Google Gemini version the platform launched with, according to Fortune, TechCrunch and TechRadar.
GenAI.mil launched in December and already counts 1.7 million active users among that 3 million-strong workforce, the three outlets report. The Gemini version remains available alongside the new additions, according to TechCrunch.
December's Gemini rollout had a rocky start: reporting by DefenseScoop, cited by Fortune, found that some workers were unsettled by an abrupt pop-up announcing the tool on their computers, with at least one senior official reportedly wondering if it was a hack.
Two vendors get a seat, a third stays out
ChatGPT Mil comes out of OpenAI's "OpenAI for Government" program and keeps data inside the department's secure environment rather than using it to train OpenAI's models, according to Fortune. TechCrunch adds that the military version is exempt from the data collection built into the consumer product.
Both Fortune and TechCrunch describe the intended use as routine, document-heavy work: administrative tasks, logistics and contracting. Fortune reports that the vast majority of military work is unclassified and deals with exactly that kind of paperwork.
Grok for Government runs through Starshield AI, the SpaceX unit that also operates the Starlink satellite network, according to Fortune and TechCrunch. Fortune reports it will offer multiple reasoning modes — auto, fast and expert. In the Pentagon's own release, quoted by both Fortune and TechCrunch, the promise is that it will let troops "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."
Both tools are cleared for controlled unclassified information at what the Pentagon calls Impact Level 5 — sensitive but not classified data — according to Fortune.
TechRadar notes the Pentagon also has separate AI agreements with Amazon Web Services, Microsoft, Nvidia and Reflection AI, part of a wider push to expand its AI capacity.
Other AI vendors appear to have taken note of Anthropic's fallout with the department. TechRadar reports that both Google and OpenAI now let the Pentagon use their models for "any lawful government purpose," a broader condition than Anthropic agreed to impose on Claude.
Notably missing from the rollout is Anthropic's Claude, the one major AI vendor with a prior Pentagon contract that didn't make this week's addition, according to Fortune, TechCrunch and TechRadar.
A dispute that already reached a courtroom
Tensions between Anthropic and the Pentagon trace back to January, when Hegseth publicly criticized AI models that "won't allow you to fight wars" — a jab later revealed to target Anthropic, amid contract talks in which the company was pushing to bar Claude's use in autonomous weapons and mass surveillance, according to Fortune.
Hegseth went on to label Anthropic a "supply-chain risk." In February, President Trump directed federal agencies to stop using the company's technology, a process that was supposed to wrap up by the end of this month, according to Fortune.
Anthropic sued over that designation. A federal judge recently ruled the Department of War acted unlawfully in labeling the company a supply-chain risk, according to Fortune and TechRadar.
The label came amid a broader push by Hegseth to fold AI into the military's daily routines. In a December memo, he wrote that he expects "every member of the Department" to use AI tools "every single day," according to Fortune. He framed the push in stark terms, writing that "the future of American warfare is here, and it's spelled AI," Fortune reports.
