Google is negotiating a deal with the Department of Defense that would allow classified use of its Gemini AI models, building on a November agreement for unclassified military use, with discussions reportedly including expanded compute capacity through Google Distributed Cloud and potentially the first classified deployment of Google’s TPU chips. Google has proposed contract language restricting its models from being used for autonomous weapons and domestic mass surveillance without human oversight, the same terms that led to the Pentagon’s dispute with Anthropic earlier this year. The talks represent a significant shift for Google, which pulled out of Pentagon AI work in 2018 following employee protests over Project Maven.
