The US Defense Department is expanding its AI offering. Through the GenAI.mil portal — open to around three million people across the defense apparatus, with 1.7 million users already on it — you can now use OpenAI’s ChatGPT and xAI’s Grok alongside Google’s Gemini. One name is still conspicuously missing from the list: Claude.
And yet Anthropic was part of the same 2025 prototype awards as the others. The reason for its absence isn’t technical, it’s a dispute over safeguard clauses. Anthropic insisted on stricter contractual limits — clauses that stop Claude from being used for mass surveillance of US citizens or in connection with autonomous weapons. The Pentagon didn’t want those limits.
That conflict turned into something concrete in February. The Defense Department labeled Anthropic a national-security “supply-chain risk” after the company refused to strip its safeguards. For a company that’s about to go public, that kind of stamp is anything but pleasant.
Then, on August 27, came the legal turn. Federal Judge Rita Lin in San Francisco ruled the Pentagon had acted illegally. The risk designation, she found, was unconstitutional retaliation — a punishment for Anthropic insisting on its terms. A clear win for the company in court.
Except: on GenAI.mil, that win changes nothing for now. Despite the ruling, Claude still isn’t part of the portal, while ChatGPT and Grok have just been added. Between a won lawsuit and actual access to a government system, there’s apparently still a fair bit of road.
For me this is one of those stories where two principles grind against each other. On one side, a provider that would rather risk a big government deal than give up its red lines on surveillance and autonomous weapons. On the other, a department that has no appetite for someone else’s conditions. You can respect Anthropic for the stance and still see that it comes at a price. The interesting question is whether the court ruling eventually lowers that price — or whether Claude simply stays the AI model that would rather say no.
Sources: DefenseScoop: Grok, ChatGPT added to GenAI.mil, The Register: Pentagon blacklisted Anthropic