The timing couldn’t be more fitting — or more cynical, depending on your perspective: while Anthropic is just now holding peace talks at the White House, Google is already actively negotiating with the Pentagon to deploy Gemini in classified networks.
What’s on the Table
According to Bloomberg and Newsweek, Google and the Department of Defense are discussing an agreement that would clear Gemini for ‘all lawful uses.’ Google has proposed its own conditions: no domestic mass surveillance and no autonomous weapons without appropriate human oversight.
That sounds remarkably similar to the exact red lines that got Anthropic locked out by the Pentagon. The difference is that Google appears to be framing it more diplomatically — or the Pentagon has become more willing to compromise after the Anthropic experience.
The Context: Google’s Defense History
For Google, this is a notable reversal. In 2018, the company shut down Project Maven — a Pentagon program for drone image analysis — after massive internal protests. Since then, Google has gradually softened its stance on military work. Gemini in classified networks would be the most significant step back into defense work yet.
OpenAI had already signed a deal with the Pentagon in February 2026 for deploying its models in classified environments.
My Take
What’s happening here is a paradigm shift. Two years ago, no major AI company wanted to be associated with the military. Today, all three big players — OpenAI, Google, and (reluctantly) Anthropic — are in talks with the Pentagon. The difference lies in the conditions: Anthropic wants clear boundaries, Google proposes similar boundaries, and OpenAI seems to have the fewest reservations. For those of us who use these models daily, this means the AI we chat with is also becoming the AI that protects — or surveils — military networks. Worth keeping in mind.
Sources: