David Robinson, an executive in OpenAI’s safety systems team, departed last week following his work on policy planning and model transparency initiatives like system cards. His exit marks another leadership change in the safety division, coming after safety lead Johannes Heidecke left earlier this year. The departure coincides with broader internal instability at OpenAI, which recently parted ways with three researchers for violating sensitive information sharing policies. CEO Sam Altman emphasized at Tuesday’s developer conference that safety and alignment must outpace model capabilities, even as Robinson had previously expressed uncertainty about whether the company's pace of change is sufficient to address challenges posed by high-performance AI models.