David Robinson, who led the writing of safety reports at OpenAI for roughly three and a half years, has resigned and is publicly warning that the company's culture is 'broken.' His essay, published in The Atlantic, was first reported by Business Insider and quickly picked up by TechCrunch, The Verge and The Guardian.
The resignation lands in a growing wave of AI safety workers leaving top labs and going public. The Verge notes that Jacob Coxon, Robert O'Callahan, Bilal Chughtai, Josh Engels and Joe Benton have all recently left Anthropic or Google DeepMind with similar warnings, making Robinson the latest and loudest voice in an expanding debate over how fast AI should be built.
David Robinson's resignation letter and what he said
Robinson says the problem goes deeper than new rules or laws. Silicon Valley, he argues, runs on 'extreme confidence' and 'perpetual sprints,' building bigger models with 'unimpeded optimism' that ignores or underestimates risk. He called himself 'something of a cliché' — the departing employee who warns the world about the product he helped ship.
In a statement responding to the essay, OpenAI spokesperson Drew Pusateri said the company continues to improve safety. Per TechCrunch, Pusateri said OpenAI pauses training or holds back models when needed, strengthens security in research and testing environments, expands work with third-party evaluators, and improves real-time monitoring to catch concerning behavior earlier in training.
- Robinson spent three and a half years at OpenAI and says he was among its longest-tenured employees.
- He wrote that OpenAI's 'trial and error' approach — which it calls 'iterative deployment' — 'guarantees periodic failures.'
