EXCLUSIVE: Deepmind Whistleblower EXPOSES Pentagon AI Contract

0:00 – 2:14

The Classified Contract That Broke the Promise

Google DeepMind, the UK-based AI division of Google, recently tore up a pledge it made years ago: no AI for weapons. Last week, executives told furious staff they were "proud" to have signed a Pentagon deal extending a $200 million contract. The new terms allow the military to use Google's AI tools for "any lawful government purpose," with Google surrendering all veto rights over classified operations. Employees responded with an open letter: "We want to see AI benefit humanity, not to see it being used in inhumane or extremely harmful ways. This includes lethal autonomous weapons and mass surveillance but extends beyond. The only way to guarantee that Google does not become associated with such harms is to reject any classified workloads. Otherwise, such uses may occur without our knowledge or the power to stop them." Their demand mirrors the stance taken by competitor Anthropic, which refused to allow Pentagon use of its AI for surveillance or autonomous weapons earlier this year. As a result, the Department of Defense labeled Anthropic a "supply chain risk" and turned to OpenAI and Google DeepMind instead. The carve-out Google workers are asking for—explicit contract language banning lethal and surveillance applications—is exactly what Anthropic insisted on, and exactly what Google refused to provide. This isn't Google's first rodeo with defense controversy. In 2017, the company signed on to Project Maven, a Pentagon initiative integrating AI into target selection, surveillance, and intelligence gathering. Workers organized protests, forcing Google to withdraw in 2018. Defense contractor Palantir then scooped up the $160 million annual contract. Project Maven was recently implicated in a US strike on a girls' school in Iran that killed over 150 civilians, including 120 schoolchildren. The system identified the school as a target.