OpenAI’s head of safety systems, Johannes Heidecke, told staff this week that he is leaving the company, according to a report from WIRED. The departure comes as OpenAI reorganizes its safety and research teams, integrating them to address what the company describes as growing coordination challenges.
In a memo to staff seen by WIRED, chief research officer Mark Chen explained the restructuring: Mia Glaese, previously VP of research and head of alignment, will take on an expanded role as VP of research and safety, overseeing OpenAI’s safety teams. Saachi Jain, who formerly led safety teams at OpenAI, will serve as interim head of safety systems, reporting directly to Glaese.
“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,” Chen wrote. “As a result, we have bigger coordination challenges around safety today than ever before.”
Heidecke joined OpenAI in 2021 as an AI safety analyst and became head of safety systems in 2024, after his predecessor Lilian Weng left to cofound Thinking Machines Lab with other OpenAI researchers. Chen thanked Heidecke for his contributions, stating, “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. We’re excited for this next chapter under Mia Glaese’s leadership across research and safety.”
Heidecke is not the only safety-focused leader to depart. Earlier this week, OpenAI’s chief futurist, Joshua Achiam, told colleagues he would leave after nine years researching safety. Additionally, OpenAI’s CEO of AGI deployment, Fidji Simo, stepped down after an extended medical leave, with Greg Brockman continuing to lead product teams and taking on go-to-market strategy.
The shakeups come as OpenAI released GPT-5.6, its most capable model to date on agentic coding tasks, according to the company. However, OpenAI acknowledged that GPT-5.6 exhibited “concerning forms of misaligned behavior,” though specific details were not disclosed.
For enterprise technology leaders evaluating AI deployment, these developments highlight the tension between rapid model releases and safety governance. Chen’s own memo acknowledged that faster training cadences and shorter release cycles are creating unprecedented safety coordination challenges. The departure of multiple safety leaders, combined with reported misalignment in the latest model, underscores the importance of independent validation for organizations integrating AI into critical business processes.
Sources:
