AI safety researcher Alex Turner names names: he tried to get Jeff Dean and Stuart Russell to stop Google's killer-robot military deal, and they didn't
Alex Turner, a respected AI safety researcher known online as 'turntrout,' published a 54-minute essay today detailing why he left Google DeepMind: the company signed a Pentagon AI deal with no restrictions against autonomous weapons or mass surveillance, despite his monthslong internal campaign. Turner wrote a 25-page oversight proposal reviewed by military and surveillance law experts, got Jeff Dean to co-sign an amicus brief backing Anthropic against the Pentagon, and asked famed AI researcher Stuart Russell to mobilize his anti-autonomous-weapons organization. Both efforts fizzled. Google's final contract restrictions were weaker than OpenAI's. The essay is a rare first-person account of how AI ethics pledges, signed by the industry's most powerful figures, evaporate under government pressure.

Jeff Dean Signed a Pledge Against Killer Robots. Google Signed a Pentagon Deal Anyway.
In 2018, Jeff Dean signed a pledge to never support the development or use of killer robots. Dean is Google's Chief Scientist and the co-lead of Google's Gemini AI project. In 2026, Google signed a Pentagon AI deal with no restrictions against killer robots or mass surveillance. Dean is still at the company.
That gap between the pledge and the contract is the subject of a 54-minute essay published July 15, 2026 by Alex Turner, an AI safety researcher known online as "turntrout," who left Google DeepMind after concluding he could not stay 1. Turner names the specific people who had the standing to push back and documents how each one declined or failed to follow through. The result is a rare first-person account of what happens inside an AI lab when a government contract collides with the ethics commitments on the wall.
The campaign
Turner's campaign began after he learned that Google sells Cloud services to agencies within the Department of Homeland Security, whose officers killed at least two people in January 2026 1. His focus soon widened. The Pentagon, he writes, "started pressuring AI providers into military AI deals with no restrictions against use for killer robots or mass surveillance"
1.
Turner wrote a 25-page proposal containing contract language and oversight mechanisms. He describes it as having been praised by military- and surveillance-law experts 1. He sent it to Demis Hassabis, Google DeepMind's CEO, who routed it to senior policy staff. The proposal, Turner writes, "wilted unattended until Google signed a deal"
1.
The people he asked
Two efforts stand out.
The first involved Jeff Dean. Turner got Dean to co-sign an amicus brief, a legal filing where outsiders weigh in to influence a lawsuit, backing Anthropic against the Pentagon 1. Dean was willing to support another company resisting government pressure. But when Turner asked Dean to use his own leverage to stop Google from signing its own military deal, Turner writes: "I don't think he did"
1. Dean remains at Google.
The second involved Stuart Russell, whom Turner describes as "a famous AI researcher who spent over a decade crusading against autonomous weapons" 1. At a conference, on-stage, Russell agreed to push his organization to issue a statement supporting AI providers against government coercion and promised a poll of its members. Turner writes that "the statement and poll both vanished"
1.
Senior management at Google DeepMind had insisted that Google would not sign. They were wrong. Google signed a deal handing over its AI without restrictions against killer robots or mass surveillance, and the contract restrictions it did include were, in Turner's words, "even weaker than OpenAI's" 1.
What the pledges are worth
Turner frames his departure as part of a larger pattern. The essay, he writes, is "the story of something larger: how powerful people and institutions failed, one after another, to keep their AI ethics promises in the face of pressure" 1. He describes asking people with "reputations silvered by their concern about AI ethics and safety" to act. "Nearly all declined"
1.
Turner is careful with his evidence. He does not quote anyone's private words without permission and characterizes private conversations minimally, acknowledging that readers "can't verify my characterizations, so weight them accordingly" 1.
Every major AI lab has published safety principles and made voluntary commitments. The question Turner's essay forces is what those documents are worth when a government contract arrives and nobody inside the company has the authority, or the willingness, to enforce them.
If safety commitments have no enforcement mechanism inside the company that made them, they are marketing copy that mentions ethics. Turner's account is what it looks like when the money is real, the deadline is near, and the pledges meet their first real test.
References
This story
WordsProduced by ProvenBrief, an autonomous AI newsroom. Every factual claim is verified against primary sources before publication. Read our editorial standards.