Openai Safety Chief Johannes Heidecke Departs Amid Team Restructuring
OpenAI's head of safety systems Johannes Heidecke is leaving the company as the AI lab merges its safety and research divisions under a unified leadership structure. The move comes as model release cycles accelerate and safety demands intensify.
OpenAI’s head of safety systems, Johannes Heidecke, has informed colleagues he is departing the company, according to an internal memo seen by WIRED. His exit coincides with a reorganization that places the firm’s safety teams under the same leadership as its research division, signaling a shift in how the AI lab manages risk as it ships models at a faster pace.
The restructuring elevates Mia Glaese, previously vice president of research and head of alignment, to a new combined role as vice president of research and safety. Saachi Jain, who has led safety teams at OpenAI, will step in as interim head of safety systems reporting to Glaese.
Details
The changes reflect OpenAI’s effort to tighten coordination between the researchers building advanced models and the teams responsible for evaluating their risks. Chief research officer Mark Chen outlined the rationale in a memo to staff this week.
- Johannes Heidecke, head of safety systems, is leaving the company
- Safety teams will now report to Mia Glaese, newly named VP of research and safety
- Saachi Jain appointed interim head of safety systems
- Reorganization aims to integrate safety and research functions more closely
- Move comes as model training cadence accelerates and release cycles shorten
Quotes
In the memo, Chen emphasized the growing pressure on safety infrastructure. “The demands on safety continue to increase — we are training models at a much faster cadence, and release cycles have come down greatly in turn,” he wrote. The statement underscores a tension familiar across the AI sector: the push to deploy capabilities quickly versus the need for rigorous evaluation.
Background
OpenAI has reorganized its safety function multiple times since its founding. The company initially maintained a distinct safety team, then created a “superalignment” group focused on long-term risks from superintelligent systems, which was dissolved last year. Glaese, who joined from DeepMind, has led alignment research — the effort to ensure AI systems act in accordance with human intent. Jain’s promotion marks her second stint leading safety systems on an interim basis.
The latest shift mirrors a broader industry trend. Rivals including Anthropic and Google DeepMind have also consolidated safety oversight under research leadership, arguing that proximity to model development enables earlier risk detection. Critics, however, warn that merging the functions could dilute independent scrutiny.
Conclusion
Heidecke’s departure and the structural overhaul arrive at a pivotal moment for OpenAI. The company is reportedly preparing new model releases while facing scrutiny from regulators, policymakers, and its own board over safety governance. With Glaese now overseeing both research and safety, the lab is betting that unified command will keep pace with its accelerating release schedule. Whether that bet pays off — or whether the loss of a dedicated safety chief creates blind spots — will become clear in the models OpenAI ships next.
Discover more from NewzQuest
Subscribe to get the latest posts sent to your email.
