OpenAI’s head of safety systems, Johannes Heidecke, is leaving the company as part of a sweeping reorganization that merges the AI firm’s research and safety teams, according to a report from Wired.
Heidecke informed staff of his departure via an internal memo seen by Wired, and Saachi Jain — who has previously led OpenAI’s safety teams — will step in as interim head of safety systems. The moves are part of a broader restructuring that places safety teams under Mia Glaese, who has been promoted to vice president of research and safety.
A Shifting Safety Structure
Heidecke joined OpenAI in 2021 and oversaw the company’s safety systems as the organization navigated an increasingly scrutinized AI landscape. His departure comes at a pivotal moment: OpenAI recently secured US government approval for GPT-5.6, its latest frontier model, and is racing to commercialize cutting-edge capabilities while managing growing concerns about responsible AI development.
Mark Chen, OpenAI’s chief research officer, told Wired that the reorganization reflects a deliberate strategy to embed safety more deeply into the model development process. “It’s important that our safety work is integrated with frontier-model development, with an earlier and more direct role in shaping key model, product and launch decisions,” Chen said.
The Reorganization in Context
The restructuring signals a shift in how OpenAI approaches safety oversight. Rather than maintaining a separate safety division, the company is weaving safety teams into its research hierarchy — a move that could accelerate decision-making but also raises questions about the independence of safety evaluations.
OpenAI’s Preparedness team — tasked with anticipating and mitigating severe risks from AI systems — remains intact. The company hired a new Head of Preparedness earlier this year, a role CEO Sam Altman described as focused on preparing for and mitigating severe risks.
The changes also follow a pattern of high-profile leadership departures at OpenAI. The company has seen several key figures leave in recent years, including former chief scientist Ilya Sutskever and alignment researcher Jan Leike, who both departed amid differences over the company’s safety culture and priorities.
What Comes Next
With Jain serving as interim head of safety systems and Glaese overseeing both research and safety at the VP level, OpenAI is effectively consolidating its safety apparatus under a unified leadership structure. Whether this approach strengthens or weakens safety oversight will likely depend on how effectively the integrated teams maintain independent judgment while embedded within the research division.
The departure of another safety leader also comes at a time when regulators around the world are scrutinizing AI companies more closely. Governments in the US, European Union, and Canada have all proposed or enacted AI safety frameworks, and the industry is under growing pressure to demonstrate that its self-regulation efforts are genuine.
For now, OpenAI continues to push forward with GPT-5.6 while working to define what responsible AI development looks like — both internally and for the broader industry.