Google’s Gemini AI model is cleared for classified use, says the Pentagon, becoming the latest AI entity to enter military service. The move would appear to provide the Pentagon with AI flexibility so that no one company’s contractual restrictions can hamper its AI use. The deal can be seen as a solution to the Pentagon’s difficulties with Anthropic’s contract terms regarding the use of the company’s AI. The Gemini deal mirrors similar arrangements the Pentagon now has with OpenAI and Elon Musk’s xAI for classified use.
Google’s move comes on the heels of a letter signed by more than 600 Google employees urging the company not to sign a deal with the Pentagon. “We want to see AI benefit humanity, not being used in inhumane or extremely harmful ways,” said the open letter signed by more than 20 directors, senior directors and vice presidents. In 2018, an employee uprising over a deal between Google and the Pentagon led to the termination of the contract.
For its part, Google says it’s proud to be a new AI recruit. But, in perhaps a balm to internal dissent, a Google spokesperson said: “We remain committed to the private and public consensus that AI should not be used for surveillance or autonomous weaponry without appropriate human oversight.”
The Google deal allows the Pentagon to use Gemini AI for what’s being described as any lawful government purpose. Critics have long noted that in the absence of any legal guidelines and the Trump administration’s reluctance to curtail AI in almost any regard, “lawful purpose” can be defined by the Pentagon.
How much input AI developers actually have regarding the Pentagon’s AI use was red-flagged by Anthropic in its court case with the Pentagon over the company’s unprecedented “supply chain risk” designation. Anthropic revealed that the company cannot manipulate its Claude AI model once it is deployed. An inability to regulate, govern or intervene in its AI tool at that stage lies at the heart of Anthropic’s dispute regarding its classified military use.
The Pentagon, meanwhile, has been adamant that no company should be telling it how to use AI. The Pentagon is asking for $2.3 billion to expand its Maven AI architecture, built by Palantir, to further integrate AI into its operations. Maven currently uses Anthropic’s AI to analyze certain aspects of intelligence. Any move by the Pentagon to terminate the relationship with Anthropic, however, is now more complicated given the company’s recent development of Mythos, reportedly the most powerful AI now in existence. The Trump administration now appears to be backing off from its fight with Anthropic.
Google’s classified deal with the Pentagon may be considered an evolution of its existing relationship with the Pentagon. Gemini AI is already in widespread use since its adoption in December, 2025 into the Pentagon’s GenAI.mil network where it automates non-classified tasks. More than 100,000 AI agents are reportedly hard at work in the Pentagon. GenAI.mil is now available across the federal government.
