David Robinson, a leader on OpenAI’s safety team, resigns and argues the company’s approach to AI development is not sufficiently careful. In an essay published by The Atlantic, Robinson says OpenAI has “thrived by trial and error,” and that its “culture is broken.” Several outlets report he left because he believes OpenAI is failing to meet the level of care he thinks is needed as it continues releasing new products.
Robinson previously led “transparency” and safety-report work tied to OpenAI’s model releases. He says the safety research and expertise should play a larger role before more powerful AI systems are developed. Outlets also highlight his call for safeguards comparable to those used in nuclear facilities—an analogy meant to stress rigorous, structured risk management rather than iterative learning after launches. Some coverage additionally references broader expert warnings about existential risk, framing his departure within ongoing debate about how quickly capabilities are advancing.
