OpenAI Safety Employee Resigns, Citing 'Broken Culture'

An OpenAI safety employee resigned, claiming the company's culture is 'broken' and calling for nuclear-level safety.

David Robinson, an OpenAI safety employee, has resigned from his position, stating that the company's 'culture is broken.' Robinson, who led the writing of safety reports for major OpenAI model releases and was among the longest-tenured employees, expressed his concerns in an editorial. He argued that the company's 'iterative deployment' approach, while leading to growth, guarantees periodic failures, with the scale of these failures increasing as systems become more capable.

Robinson's departure echoes previous warnings from others in the AI industry. He emphasized that frontier AI companies should operate with the rigorous safety standards of nuclear power plants or busy airports, incorporating layers of redundancy and careful planning. He noted a lack of colleagues with experience in such high-stakes safety environments during his tenure. OpenAI spokesperson Drew Pusateri responded, stating the company is continuously improving its safety measures, strengthening security, and training models for responsible task completion.

Beyond cultural changes, Robinson called for a deeper examination of AI alignment, acknowledging it might sound 'touchy-feely' but is critical as current measures of AI systems matching human values are 'coarse.' He warned that the more the industry allows models to grow without solving these problems, the more dangerous the situation becomes. Robinson's resignation and warnings highlight ongoing debates about AI safety and the internal practices of leading AI companies.

Find the right AI tool for your business

Chat with Insta and get matched to the right tool in seconds.