The head of OpenAI's security system leaves the job and the senior management leaves the bandwagon to add another member

Zhitong Finance App learned that, according to reports, Johannes Heidecke (Johannes Heidecke), head of OpenAI's security systems, told his colleagues this week that he is leaving the industry-leading artificial intelligence (AI) company. This is the latest in OpenAI's recent wave of key executive departures. What is behind it is the company's security team restructuring to cope with the ever-increasing pace of model iteration.

The report quoted an internal memorandum as saying that OpenAI's chief research officer Mark Chen (Mark Chen) has announced that the company's security team will report to Mia Glaese (Mia Glaese), vice president of research and head of alignment in the future. Glace's role has been significantly expanded in this adjustment to become vice president of research and safety, unifying the two major sectors of research and safety. Meanwhile, Saachi Jain (Saachi Jain), who previously led the security team, will be the head of the temporary security system and will report to Glace.

Chen explained the reason for this restructuring in the memorandum, saying that safety requirements are continuing to increase, the pace of model training has been greatly accelerated, and the release cycle has also been significantly shortened. This has made coordination challenges surrounding safety issues more serious than ever before. In a statement, he said, “We are grateful to Johannes for his contributions to OpenAI. Integrating security work with cutting-edge model development is critical to enable security to play an earlier and more direct role in key model, product, and launch decisions. We're excited for Mia Glace to lead a new chapter in research and security.”

Heideck joined OpenAI in 2021, initially as an AI security analyst, and took over this position after Lilian Weng (Lilian Weng), the head of security systems, left to found Thinking Machines Lab before 2024. His departure coincided with OpenAI trying to launch increasingly powerful models. This week, the company just released GPT-5.6, its most capable model for intelligent programming tasks, but compared to the previous model, the government acknowledged that GPT-5.6 showed worrying alignment issues.

Heideck's departure is not an isolated incident; OpenAI's research and product leadership are undergoing chain changes. Earlier this week, Joshua Achiam (Joshua Achiam), the chief futurist who has been doing safety research at the company for nine years, also informed colleagues that he is leaving the company. Additionally, OpenAI's CEO Fidji Simo (Fidji Simo), who is responsible for the AGI deployment, also informed employees this week that he will step down after a long period of sick leave. The company said Greg Brockman (Greg Brockman) will continue to lead the product team and take over market entry strategies outside of his original responsibilities.