OpenAI safety lead quits, calling company culture broken
OpenAI safety culture is under fire as David Robinson exits over rushed AI safeguards and warns labs need stronger redundancy before faster launches.
OpenAI safety lead David Robinson has left the company after 3.5 years, saying in an Atlantic essay that the organization’s culture is broken and that AI safety work is moving too fast.
Robinson helped oversee safety reports for multiple frontier model launches and contributed to OpenAI’s Preparedness Framework, which is used to guide model safety decisions. In his view, AI labs should operate with stronger redundancy and more rigorous safeguards, similar to systems used in aviation or nuclear facilities.
His departure comes as OpenAI continues to face scrutiny over safety, internal staffing changes, and model launches that have raised concerns about whether product speed is outweighing caution. The article also notes that Robinson’s criticism echoes earlier warnings from former OpenAI staff who said safety was losing priority.