OpenAI safety staffer quits, calls for stronger safeguards

David Robinson, who says he wrote the safety reports that accompanied OpenAI’s major launches, has left the company and argued that its culture is built for speed rather than the caution increasingly powerful AI demands. His complaint is not simply that a rule is missing: OpenAI’s approach finds problems in real use and tightens safeguards afterward, a model he says makes some failures unavoidable. Robinson wants frontier AI labs to operate more like aviation or nuclear power, with backup protections and deliberate planning designed to keep ordinary human mistakes from becoming disasters. OpenAI disputed the broader characterization, saying it pauses training or holds back models when it needs to slow down. The departure sharpens a central question for the industry: can safety processes mature as quickly as the systems they are meant to control?