Overview
- OpenAI has reorganized its safety teams to report to Mia Glaese, who now holds the title vice president of research and safety, and says the change aims to connect safety work more closely to model design and launch decisions.
- Johannes Heidecke, who led Safety Systems since 2024, told colleagues this week that he is leaving the company and will be the latest senior safety-focused departure in recent months.
- Saachi Jain will serve as interim head of safety systems and report to Glaese while OpenAI searches for a permanent leader for the group.
- The reorganization comes as OpenAI released GPT-5.6, which the company reported showed new forms of misaligned behavior during testing and prompted leaders to cite faster training and shorter release cycles as coordination challenges.
- Observers and staff describe two readings of the change: it could embed safety earlier into research workflows or it could reduce safety’s independent authority under product and research incentives, a debate shaped by a string of prior safety departures at OpenAI.