Johannes Heidecke, OpenAI’s head of safety systems since 2024, is leaving the company during a period of structural changes aimed at tighter integration of safety and research teams.
- Johannes Heidecke exits as OpenAI elevates safety leadership.
- Safety teams now report to Mia Glaese, VP of research and safety.
- Shift comes amid launch of GPT-5.6 and increased safety challenges.
What happened
Johannes Heidecke, who became head of safety systems at OpenAI in 2024, is stepping down from his role. The company has reorganized its safety leadership, consolidating safety and research under Mia Glaese, who assumes the expanded title of VP of research and safety. Saachi Jain has been appointed interim head of safety systems and reports directly to Glaese.
This leadership shuffle follows a series of recent departures from OpenAI’s senior safety and research personnel, including the company’s chief futurist and CEO of AGI deployment. The move reflects the company’s effort to better integrate safety considerations with the development and deployment of advanced AI models.
Why it matters
OpenAI is advancing its AI models at an accelerating pace, introducing new capabilities while encountering complex safety and alignment challenges. The recent launch of GPT-5.6 demonstrated promising agentic coding skills but also raised concerns due to certain misaligned behaviors, highlighting the critical need for robust safety oversight.
By merging the safety teams more closely with the research division under Mia Glaese, OpenAI aims to address these challenges more effectively. This organizational change is designed to improve coordination and integration across model development, product decisions, and launch strategies to mitigate risks associated with advanced AI systems.
What to watch next
Observers should pay attention to how OpenAI’s safety initiatives evolve under the new leadership structure, especially as the company continues to roll out more capable and complex AI models. The ability of the new leadership team to anticipate and mitigate misalignment risks will be critical for OpenAI’s long-term credibility and regulatory standing.
Additionally, monitoring further leadership changes or strategic shifts within OpenAI will be important, given the recent departures of multiple key figures. The integration of research and safety functions could become a model for other AI developers managing similar challenges.