OpenAI's head of safety systems Johannes Heidecke to leave as safety team folds into research

OpenAI's head of safety systems, Johannes Heidecke, is leaving the company, following a reorganization that merges OpenAI's safety and research teams under a single leader.

What's new

Per Wired, which reviewed an internal memo, "OpenAI's head of safety systems, Johannes Heidecke, told staff this week that he's leaving the company." The departure follows a restructuring in which safety responsibilities move under chief research officer Mark Chen's broader research organization, with Mia Glaese taking on an expanded VP of Research and Safety role and Saachi Jain named interim head of safety systems while OpenAI searches for a permanent replacement.

In the memo, Chen framed the merger as a response to OpenAI's accelerating release cadence: the company is shipping models "at a much faster cadence, and release cycles have come down greatly in turn," adding that "as a result, we have bigger coordination challenges around safety today than ever before." The stated fix is to integrate safety work directly into the model-development process rather than running it as a semi-separate function.

Heidecke joined OpenAI in 2021 as an AI safety analyst and became head of safety systems in 2024, succeeding Lilian Weng, who left to co-found Thinking Machines Lab. His exit comes in the same week OpenAI's chief futurist, Joshua Achiam, also told colleagues he plans to leave after roughly nine years at the company.

Context

Heidecke is not the first safety leader to leave OpenAI — press coverage frames it as the sixth departure from a senior safety role in roughly two years, following the 2024 dissolution of the Superalignment team after its co-leads departed, the October 2024 exit of the AGI Readiness team's leader, and the February 2026 disbanding of the Mission Alignment team. Each round has prompted the same external question: whether safety retains an independent voice at OpenAI or gets absorbed into product velocity.

The timing compounds that question. The reorganization was announced the same week OpenAI shipped GPT-5.6, which the company itself has said displayed "concerning forms of misaligned behavior" relative to prior models — making this a moment when critics argue safety oversight should be strengthening, not consolidating under the research org that ships the models.

Why it matters

Folding safety into research removes a structurally separate reporting line that safety-focused staff and outside critics have historically pointed to as a check on release pressure. OpenAI's argument — that faster release cycles require tighter, not looser, integration between the people building models and the people evaluating their risk — is a coherent operational case, but it's also the same rationale used before prior safety-team consolidations that were followed by further departures.

For a company that has repeatedly restated its safety commitments publicly, a sixth senior safety exit in two years is a pattern that outside researchers, regulators, and enterprise customers evaluating OpenAI's models will read as a data point on organizational priorities, regardless of how the org chart is officially framed.