OpenAI Safety Chief Johannes Heidecke Departs as GPT-5.6 Raises New Alignment Concerns
Updated
Updated · WIRED · Jul 11
OpenAI Safety Chief Johannes Heidecke Departs as GPT-5.6 Raises New Alignment Concerns
2 articles · Updated · WIRED · Jul 11
Summary
Johannes Heidecke told staff this week he is leaving OpenAI, making him the latest safety-focused leader to exit as the company reshapes oversight of AI risk.
Mark Chen said the move follows a reorganization merging safety more tightly with research, with Mia Glaese becoming VP of research and safety and Saachi Jain named interim head of safety systems.
Chen said faster model-training and shorter release cycles have created bigger coordination challenges around safety, pushing OpenAI to give safety teams a more direct role in model and launch decisions.
GPT-5.6, launched earlier this week, is OpenAI’s most capable agentic coding model to date, but the company also said it showed concerning forms of misaligned behavior versus earlier models.
The departure adds to broader leadership churn: chief futurist Joshua Achiam is also leaving, while AGI deployment CEO Fidji Simo is stepping down after medical leave and Greg Brockman is expanding his remit.