OpenAI's head of safety systems is out, and the team he led no longer stands on its own.
Johannes Heidecke is leaving OpenAI following an internal restructuring that dissolved the separation between safety and research. Chief Research Officer Mark Chen announced in a staff memo that safety teams would now report to Mia Glaese, whose title was expanded to reflect the broader mandate. The move effectively ends safety's status as an independent function at the company.
The timing is notable. OpenAI has faced sustained criticism for deprioritizing safety work as it races to ship products, and a series of high-profile departures from its safety team over the past two years has fed that narrative. Folding safety under research could reflect a genuine belief that the two disciplines belong together - or it could mean safety now competes for attention inside a team whose primary job is building faster. Those are very different outcomes that look identical on an org chart.
OpenAI has insisted before that safety and capability research are complementary, not in tension. Regulators and former employees have been less convinced.