Pentagon’s Secret AI Deal: Google Demands Limits on Surveillance and Autonomous Weapons

The Pentagon‘s working on a secret deal with Google to use its Gemini AI models for military purposes. The agreement would let the military use the AI for all lawful tasks. But Google’s setting strict rules on how the technology can be used.

Google’s refusing to let the military use its AI for mass surveillance of American citizens. It’s also blocking the use of fully autonomous weapons without human control. These conditions mirror terms that OpenAI already negotiated with the Pentagon. The rules come from ethical standards that Silicon Valley tech companies have adopted.

Google’s Public Sector division is leading the talks. The goal is to win major defense contracts. The deal would involve deploying custom AI chips, called TPUs, in secure military settings. Neither Google nor the Department of Defense has publicly confirmed the discussions.

Anthropic, which makes the Claude AI model, took a different path. It refused similar terms and rejected rules around autonomous weapons and surveillance. The Pentagon then labeled Anthropic a supply chain risk. That led to a six-month phase-out of Claude from government systems.

Anthropic sued the government, citing First Amendment and due process violations. A federal judge called the Pentagon’s designation “Orwellian” but still upheld the ruling.

OpenAI stepped in after Anthropic’s exit. It partnered with the Pentagon and revised its contract to include ethical restrictions. CEO Sam Altman pushed for those same terms to apply to all AI vendors. OpenAI faced some criticism for quickly filling the gap left by Anthropic.

Google’s already built a strong relationship with the Defense Department. It won a DoD AI pilot worth up to $200 million last July alongside Anthropic, OpenAI, and xAI. Gemini became the first model added to the unclassified GenAI.mil platform in December. Google Distributed Cloud achieved IL6 and Top Secret accreditation in 2025, authorizing it to handle Secret and Top Secret classified workloads.

Google’s also targeting $2 billion in defense bookings between 2025 and 2027. This deal marks a shift for Google, which once pulled back from military AI work. The company’s now pushing deeper into defense. The Pentagon also sees AI as a key tool to speed up administrative tasks and reduce operational costs. It’s a sign of growing competition among AI firms for Pentagon contracts.