Johannes Heidecke, the head of safety systems at OpenAI, has announced his departure from the artificial intelligence powerhouse. His exit coincides with a significant internal reorganization designed to dissolve the traditional barriers between the company’s dedicated safety division and its broader research operations.
Under the new structure, OpenAI is merging its safety and research teams into a unified entity. Chief Research Officer Mark Chen informed staff that Mia Glaese, currently the Vice President of Research and Integration, will step into the expanded role of Vice President of Research and Safety. This move is intended to streamline the development-to-deployment pipeline.
Saachi Jain, who previously held a leadership position within the safety group, will serve as the interim head of safety systems reporting directly to Glaese. This consolidation marks a definitive shift in how the creator of ChatGPT manages the operational risks associated with its frontier models, moving away from an independent oversight model.
The restructuring follows a tumultuous period for OpenAI, characterized by the high-profile departures of key safety advocates including co-founder Ilya Sutskever and Jan Leike. These exits have fueled an industry-wide debate over whether the company is prioritizing commercial speed over long-term AI alignment and safety protocols.
By embedding safety experts directly within the development teams, OpenAI aims to accelerate the release of its next-generation models while maintaining its competitive edge against rivals like Anthropic and Google. However, the centralization of power under research-focused executives may raise concerns among regulators regarding the independence of internal safety audits.