OpenAI's head of safety systems, Johannes Heidecke, is leaving following a company restructure that places safety under research leadership. The move reflects OpenAI's continued integration of safety into frontier AI development despite ongoing senior leadership churn.
- Safety team now reports to VP of Research and Safety, Mia Glaese
- Johannes Heidecke departs after heading safety since 2024
- Recent restructuring follows multiple senior safety leadership changes
What happened
OpenAI announced an internal restructuring that merges its AI safety team with its core research division. Johannes Heidecke, the head of safety systems since 2024, is leaving the company. Mia Glaese’s role was expanded to Vice President of Research and Safety, consolidating safety oversight under her leadership. An interim head, Saachi Jain, has been appointed while a permanent replacement is sought.
This marks the second time in less than two years that OpenAI has folded its safety function under research leadership. The change follows previous dissolutions of key AI safety teams including the Superalignment and Mission Alignment groups, along with departures of other senior safety officials. Chief Research Officer Mark Chen emphasized that the company wants safety work integrated deeply into AI model development and launch decisions.
Why it matters
OpenAI’s decision to embed safety inside research aims to ensure safety considerations influence AI development from the earliest stages rather than acting solely as a final checkpoint. This approach could enhance coordination between safety experts and product teams, potentially accelerating efforts to align models with human values and prevent harmful outcomes.
However, critics caution that merging safety functions under research leadership may reduce the independence and authority of safety teams. This structural shift could make it more difficult for safety experts to challenge or delay product launches when risks arise. The leadership turnover also raises concern over continuity and organizational focus on long-term AI safety aims as OpenAI scales its capabilities.
What to watch next
How OpenAI recruits a permanent head for safety systems and whether that person can assert sufficient influence within the merged research-safety division will be key to the company’s safety trajectory. Observers will also watch for the impact of more integrated workflows on timely identification and mitigation of novel AI risks.
In parallel, OpenAI’s regulatory and public relations environment is increasingly complex, with major state investigations into privacy, advertising, and internal governance practices underway. These external pressures may affect OpenAI’s strategic priorities around transparency and risk management in AI deployment.