OpenAI Merges Safety and Research Teams as Head of Safety Steps Down
OpenAI has folded safety into research again. Its head of safety is leaving.
Johannes Heidecke is departing after two years as head of safety systems at the company, according to a report by Wired. Following an internal restructuring, OpenAI will now merge its safety and research teams under a single leader. Mark Chen, Chief Research Officer, announced this change in a memo to staff, with Mia Glaese taking on the expanded role of VP of Research and Safety.
Saachi Jain has been appointed as the interim head of safety systems while OpenAI searches for Heidecke’s permanent replacement. This is the second time in less than two years that OpenAI has integrated its safety organization into a structure reporting directly to a research lead, reflecting a shift in priorities within the company.
Heidecke’s Tenure and Recent Departures
Heidecke, who joined OpenAI in 2021 as an AI safety analyst, played a crucial role in model alignment, rule-based reward systems, and the company’s preparedness evaluations for potentially dangerous model capabilities. His work was instrumental in ensuring that safety considerations were integrated into key model, product, and launch decisions.
OpenAI has seen several senior safety figures leave or be reorganized over the past two years, including:
- Ilya Sutskever and Jan Leike, co-leads of the Superalignment team, who departed in May 2024.
- Miles Brundage, leader of the AGI Readiness team, resigned in October 2024.
- The Mission Alignment team, formed as a successor to Superalignment, was disbanded in February 2026 after 16 months.
- Fidji Simo, OpenAI’s chief of applications, stepped down this month citing medical reasons.
Integration vs. Structural Independence
Glaese’s expanded title signifies that OpenAI wants to maintain a strong emphasis on safety even under the new structure. The company recently launched a Safety Fellowship program inviting external researchers to conduct independent safety and alignment work at the lab. Chen justified this move by stating that embedding safety within research ensures its involvement in model decisions from the outset, rather than as a final check. Critics, however, argue that a safety team reporting inside research has less structural independence and may have limited ability to delay or block a product launch.
Navigating External Scrutiny
Heidecke’s departure comes at a time when OpenAI is facing intensifying external scrutiny from state attorneys general, with 42 AGs investigating the company.