The Pentagon picked sides in the AI race this week, and the big winner is Google. The big loser is Anthropic.
The Defense Department's AI chief told CNBC that Google's Gemini model is now running on classified projects, while Anthropic is locked out of Pentagon contracts during a court fight.
Pentagon AI chief Cameron Stanley confirmed the DOD is expanding its use of Gemini, while also working with OpenAI and other vendors.
"Overreliance on one vendor is never a good thing," Stanley said. "We're seeing that, especially in software."
A person familiar with the deal told CNBC that Google (GOOGL +0.11%) is using its latest model on classified work, with The Information first reporting the deal.
Stanley said Gemini is "saving thousands of man hours, literally thousands of man hours on a weekly basis" for U.S. warfighters.
The Pentagon dropped Anthropic about two months ago, calling the company a supply chain risk, and Anthropic sued.
The legal fight has split decisions. A federal appeals court in Washington, D.C. earlier this month denied Anthropic's request to block the blacklist while the lawsuit plays out, while a judge in San Francisco granted Anthropic an injunction in a separate case that bars the Trump administration from banning Claude across government.
The result: Anthropic is out at the DOD but can still work with other federal agencies during the fight.
A DOD spokesperson confirmed the Pentagon is not currently working with Anthropic. President Trump told CNBC last week it's "possible" a deal gets done that lets Anthropic models back in.
Stanley said Anthropic's Mythos rollout earlier this month was a wakeup call, since the company released the powerful model to a limited number of companies, citing advanced cyber capabilities and the risks they posed.
The DOD is "taking this very seriously," Stanley said, so it can prepare for "a whole raft of AI-enabled capabilities."
He used a Thanksgiving analogy to describe how the Pentagon picks tools: "You don't cook a Thanksgiving turkey in the microwave."
The deal isn't sitting well inside Google, where more than 700 employees signed a letter to CEO Sundar Pichai this week asking the company to reject classified workloads.
The letter said staff don't want the technology used in "inhumane or extremely harmful ways."
The next signal is whether Trump opens the door back up for Anthropic at the Pentagon, and whether Google's internal pushback grows. Both shape who gets the long-term defense AI dollars.
