Two rulings, one platform, one absence
On August 31, 2026, the Pentagon added OpenAI's ChatGPT Mil and xAI's Grok for Government to GenAI.mil, its internal portal for giving Department of Defense personnel access to commercial artificial intelligence (AI) chat tools inside a secure government environment. Anthropic's Claude, the model the Pentagon relied on almost exclusively for over a year before cutting ties with the company, was not among them - even though a federal judge had ruled four days earlier that banning Anthropic was unlawful.
What GenAI.mil is
GenAI.mil launched in 2025 as a way to let Defense Department staff use frontier AI chat models without routing sensitive government material through the ordinary consumer versions of those products, which retain user data under standard commercial terms. It started with Google's Gemini as the only option. Grok for Government was added quietly in late December 2025, and ChatGPT Mil followed on August 31, 2026. Both new additions are accredited to handle Controlled Unclassified Information at Impact Level 5, a Defense Department classification that covers sensitive but unclassified material such as personnel records and contracting data, one tier below fully classified systems.
The portal is meant for the department's least glamorous but highest-volume work: paperwork, logistics planning, contracting, market research, and supply-chain coordination for roughly 3 million uniformed, civilian, and contractor personnel. As of the August 31 announcement, 1.7 million of them had already created accounts, putting adoption at more than half the eligible workforce well before the rollout was complete.
Defense Secretary Pete Hegseth framed the rollout as effectively mandatory, telling staff "the future of American warfare is here, and it's spelled AI" and that he expects every member of the department to log in, learn the tools, and incorporate them into daily workflows immediately. In a statement, OpenAI's Joe Larson said ChatGPT on GenAI.mil brings a familiar commercial experience into the department's secure environment for unclassified work, while a Pentagon statement described ChatGPT Mil as tailored to "warfighter needs."
How Anthropic ended up excluded
Anthropic was not always on the outside. The company held an active Pentagon contract worth up to $200 million, in the same range OpenAI and xAI's Starshield division each received for their own deals. For a time, Claude was reportedly the only model available on the Pentagon's classified network.
The relationship broke down over how the models could be used. In January 2026, Hegseth publicly criticized AI models that, in his words, "won't allow you to fight wars." The Trump administration wanted contract terms allowing Claude to be applied to "all lawful purposes," including autonomous weapons systems and mass surveillance. Anthropic declined, citing its own safety commitments, and in February 2026 the administration directed agencies to stop using Anthropic's technology, formally labeling the company a supply-chain risk - a designation the government has historically reserved for firms tied to foreign adversaries, not domestic vendors it has a policy dispute with.
Anthropic sued in March 2026, arguing the designation was retaliation for its refusal to drop its safety guardrails, in violation of its First Amendment rights. On August 27, 2026, U.S. District Judge Rita Lin agreed. In a 59-page order, she found the Pentagon's blacklisting of Anthropic was, in her words, illegal and baseless, ruling it violated both the First Amendment and the due process clause of the Fifth Amendment.
Four days after that ruling, the Pentagon expanded GenAI.mil with two more AI vendors and left Claude off the list.
What it means in practice
A court ruling that a ban was unlawful does not automatically restore a vendor to a government platform; enforcement and appeals can take much longer than the ruling itself, and nothing in the available reporting indicates the Pentagon has said what it plans to do next about Claude's status. For the roughly 3 million people GenAI.mil is built for, the practical result today is that Gemini, ChatGPT Mil, and Grok for Government are the options on the menu, regardless of how the legal dispute resolves. For Anthropic, the ruling is a significant legal win, but it has not yet translated into military users having access to Claude again - a gap worth watching for whether it closes, and how the Pentagon justifies the delay if the ruling stands unappealed. For OpenAI and xAI, the timing is also notable: both secured expanded government access in the same window Anthropic was fighting to get back in, reinforcing that willingness to accept the department's usage terms, not just model quality, is shaping who gets a seat on the platform.
