Inside OpenAI’s Latest Safety Lead Exit: Is 'Iterative Safety’ A Broken Model?

David Robinson, a senior OpenAI safety employee, resigned from the company and publicly criticised its approach to AI safety, stating that OpenAI's culture is broken and that AI firms "were not being careful enough." Robinson had worked at OpenAI for around three and a half years.

It describes Robinson's resignation as the latest exit of a safety lead at OpenAI, and poses the question of whether the iterative safety model is broken. Robinson's essay attributes the problem to the company's culture rather than to any single product or release, and his criticism is aimed at AI firms generally as well as at his former employer.

The report frames the episode as a question about whether "iterative safety" remains a workable model, using the term in its headline. Robinson's departure came after around three and a half years at the company, a period the report does not break down further. The excerpt does not include any response from OpenAI to the essay, and it does not name any other safety employee.