OpenAI safety systems lead Johannes Heidecke is leaving

In short: OpenAI’s head of safety systems, Johannes Heidecke, told staff he is leaving, and the company is reshuffling who leads its safety work.

What happened

WIRED reports that Johannes Heidecke, OpenAI’s head of safety systems, informed employees this week that he is leaving the company. Safety systems work is meant to reduce harmful or unwanted behavior from AI models (like adding guardrails to a powerful tool).

The departure follows an internal reorganization that aims to more closely connect OpenAI’s safety teams with its research teams. In a memo seen by WIRED, OpenAI chief research officer Mark Chen said the safety teams will now report to Mia Glaese. She is OpenAI’s head of “alignment,” which means work that tries to make AI follow human goals and rules. Glaese will take an expanded role as VP of research and safety.

Chen also wrote that Saachi Jain will become interim head of safety systems and will report to Glaese. Heidecke joined OpenAI in 2021 as an AI safety analyst and became head of safety systems in 2024.

The change comes as OpenAI continues to release new models more frequently, according to Chen’s memo. WIRED notes OpenAI recently launched GPT-5.6, and the company said it saw “concerning forms of misaligned behavior” compared to earlier models. The report also mentions another recent departure, OpenAI safety-focused leader Joshua Achiam.

Why it matters

As AI tools are used in more places, from writing to coding, leadership changes in safety teams can affect how quickly and carefully new features are released. For regular users, this can shape how reliable the tools feel and how well they avoid risky outputs.

Source: Wired