Google DeepMind researcher resigns over Pentagon deal, explains why in 2,000-word note

A Google DeepMind researcher has quit over the company's decision to sell its AI to the Pentagon, and he's not going quietly. Alex Turner, who spent more than two years on AI safety at DeepMind, resigned in June and this week published a 2,000-plus word account titled "Why I Left Google DeepMind," laying out months of internal lobbying that ultimately failed. "I resigned from Google DeepMind because it broke its founding promise by selling AI to the military without restrictions against killer robots or mass spying," Turner wrote on X, where the thread has since crossed 750,000 views. His central complaint is simple: Google signed a classified deal letting the Pentagon use its Gemini AI for "any lawful government purpose," with no binding limits on autonomous weapons or surveillance. Google DeepMind's 2014 military promise, broken by the Pentagon AI deal When Google acquired DeepMind in 2014, it came with a pledge that the lab's technology would never be used for military purposes. In 2018, DeepMind and its founders signed a pledge against lethal autonomous weapons. Turner argues those commitments have quietly eroded—Google stripped the weapons and surveillance prohibitions from its AI principles in early 2025, in a post co-authored by CEO Demis Hassabis. Turner says his campaign began in January after DHS agents killed two people, which led him to Google's Cloud contracts within the immigration enforcement supply chain. It widened when the Pentagon pressured Anthropic to drop its own red lines against autonomous weapons and AI profiling, threatening to designate the company a supply-chain risk. Turner feared Google would fold under similar pressure. It did. The DeepMind researcher who lobbied Google executives and got left on read This wasn't a quiet exit. Turner emailed Sundar Pichai, Hassabis, and Cloud chief Thomas Kurian. He had lunch with Chief Scientist Jeff Dean, organized a petition that drew over 250 DeepMind signatures, and wrote a 25-page framework of contract language and oversight mechanisms that a military-law expert called "actually pretty good." He even cold-messaged Hassabis directly after his emails went unanswered. Dean did sign an amicus brief backing Anthropic against the Pentagon, a move Turner credits as genuinely important—it reportedly made some Pentagon officials wary of relying on Google. But Turner says Dean's leverage left no mark on the final deal, and Dean remains at the company despite his 2018 pledge. Turner's framework, routed to senior policy staff at Hassabis's instruction, was never evaluated. "They left the message on read," he wrote. Where the AI ethics establishment stayed silent on Google's Pentagon deal Some of Turner's sharpest criticism lands on the AI safety community itself. At an IASEAI conference in Paris, leadership announced a member vote on a statement supporting Anthropic. A near-unanimous show of hands backed it. Then the vote disappeared. Turner also lobbied prominent researchers Yoshua Bengio and Stuart Russell—the latter a decade-long crusader against autonomous weapons—but neither made a public statement. Google, for its part, has said API access to its commercial models with industry-standard terms is a responsible approach to national security. Turner isn't buying it. He's currently unemployed, declined outreach from OpenAI's safety team, and says he's working on independent AI safety research while he figures out his next move.