AI Policy Wiki
Dashboard

Alex Turner

medium confidence · updated 2026-07-25

Independent AI alignment researcher, formerly a research scientist at Google DeepMind, who resigned on 9 June 2026 over Google's classified Pentagon AI contract. Organized internal opposition and drafted a 25-page alternative contract framework; argues that voluntary pledges and insider influence failed as governance mechanisms and that binding structures and legislation are required.

Alex Turner is an independent AI alignment researcher. He worked as a research scientist at Google DeepMind until June 9, 2026, and previously worked for several years in Stuart Russell's lab.

Opposition to the Pentagon contract

Turner publicly criticized Google's April 2026 "any lawful government purpose" agreement with the Department of Defense alongside fellow DeepMind research scientist Andreas Kirsch (Source: transformernews.ai). He organized a petition signed by more than 250 DeepMind employees and submitted a 25-page alternative framework — contract language plus oversight mechanisms, requiring human control over targeting — to Demis Hassabis (Source: storyboard18.com). He resigned when the deal was signed.

Positions

In "I tried to stop Google DeepMind's Pentagon deal. Then I quit." (Transformer, July 2026), Turner sets out three arguments.

On the commitments. He holds that Google's February 2025 removal of the weapons and surveillance prohibitions from its AI principles cannot be reconciled with Hassabis's subsequent statement that "nothing's changed about our principles" — "Both statements cannot be true" — and that the Pentagon contract, by signing away Google's ability to restrict use cases, appears to violate the company's 2018 pledge against lethal autonomous weapons.

On what a pledge obliges. Turner argues a pledge "binds, even when it is personally costly," and gives three options he regards as honest for a signer whose employer crosses it: explain how staying is consistent with the pledge, explain why you no longer hold it, or quit. He treats credibility as the stake — a signer who stays teaches counterparties that "these safety people will not act, even at their own brightest line," devaluing the next commitment.

On governance design. His general conclusion is that "society cannot rely on ethics-motivated people standing firm. They just don't do it reliably. We need structures: binding contracts, independent auditors, and review bodies that cannot be quietly dissolved," and eventually legislation. He offers Google DeepMind's trajectory as the test case for the seat-at-the-table theory Hassabis described adopting at acquisition, and argues the classified deal — which produced "zero binding concessions" at what Turner calls the clearest red line the Gemini project will face — should settle it: "we should at least stop telling ourselves that a seat at the table works."

He extends the same criticism outward, to Stuart Russell's undelivered statement and member poll on government coercion of AI providers, and to Global Call for AI Red Lines signatories including Yoshua Bengio and Geoffrey Hinton, whom he describes as publicly silent once the counterparty was the US government.

Relationships