According to Business Insider, Johannes Haidk, the head of OpenAI's safety systems team, is about to leave. He is the latest key member to depart from OpenAI's safety and alignment team in recent years. AI analyst Andrew Chen even used a reference from "Harry Potter" to describe the phenomenon of frequent changes in OpenAI's safety leadership as a "curse."

openai

Frequent changes in safety leadership and public doubts about company's safety attitude

Haidk joined OpenAI in 2021, taking over the team after the previous safety lead, Weng Li, left. He was responsible for risk assessments and mitigation measures before model releases. OpenAI stated that Mia Glazer, Vice President of Research and Safety, will take over, aiming to integrate safety work more deeply into all research teams. However, OpenAI has struggled to retain senior safety leaders. Some former employees have publicly questioned whether the company truly values safety.

Co-founder Jan Leike left in May 2024, stating openly that OpenAI prioritizes product launches over ensuring safety, saying, "Safety culture and safety processes have been pushed to a secondary position by flashy products." Braden Kuo, a senior advisor on AGI preparation, warned: "OpenAI is not ready, nor are any other cutting-edge labs, and the entire world is also not ready."

Eight core members have left in succession

Among those who have left the key safety teams in recent years are co-founder Ilya Sutskever, who later founded Safe Superintelligence; co-founder Jan Leike, who then joined Anthropic to continue alignment research; Vice President of Research and Safety Weng Li, who left after nearly seven years at OpenAI in November 2024. Others include the head of model policy and safety research, Walon, the head of risk preparedness team, Madrid, the chief futurist, Akiyam, and the head of the safety systems team, Haidk. Both Walon and Leike chose to join the competitor Anthropic, a trend that has continued to raise questions about OpenAI's commitment to safety.