OpenAI safety leader David Robinson has resigned after three and a half years, warning that the company’s culture is broken and questioning whether current safeguards can keep pace with increasingly capable AI.