OpenAI Confronted with a Talent Drain in AI Security

A Wave of Departures at OpenAI

OpenAI, one of the major players in the field of artificial intelligence, is facing a series of significant departures within its security teams. The most recent is Johannes Heidecke, who held the position of head of security systems. His departure adds to an already long list of security leaders who have left the company, raising questions about OpenAI's commitment to developing AI safely. These departures come as the company reorganizes its security and research efforts under the leadership of a new head.

Heidecke is not the only one leaving OpenAI. Several other leaders in security strategy have also decided to depart as the company undertakes a reorganization of its security and research efforts. This reorganization aims to centralize these efforts under a single leadership to improve the efficiency and coherence of security initiatives. This strategic decision comes at a time when OpenAI is reconsidering the relationship between its research and security efforts.

Reorganization Under New Leadership

Mark Chen, the head of research at OpenAI, expressed gratitude for Johannes Heidecke's contributions and announced that Mia Glaese would now take the reins as vice president of research and security. Under her leadership, OpenAI hopes to integrate security more deeply into its research teams, a move that could enhance the synergy between these two critical areas. This reorganization is seen as an attempt to better align research and security efforts, recognizing that one cannot be effective without the other.

This reorganization occurs as OpenAI reevaluates the relationship between its research and security efforts. A company spokesperson emphasized the importance of understanding the underlying capabilities of models to make sound security decisions, and vice versa. This integrated approach is essential to ensure that technological developments do not compromise safety.

OpenAI's Mission Under Scrutiny

OpenAI's stated mission is to ensure that general artificial intelligence benefits all of humanity. Security, defined as the practice of maximizing the positive impacts of AI while mitigating the negatives, is therefore at the heart of this mission. However, the departure of several security leaders highlights internal tensions regarding the priority given to security versus product development. Some departing employees have publicly questioned OpenAI's commitment to this crucial work.

Jan Leike, a former co-director of OpenAI's superalignment team, left the company in May 2024, publicly criticizing OpenAI's approach. He claimed that the company prioritized launching attractive products at the expense of security processes, a critique that resonates with the concerns of many other departing employees. Leike pointed out that the culture and security processes had been sidelined, which is alarming for a company whose mission is to ensure AI safety.

Departures Raise Questions

Dario Amodei, co-founder of Anthropic and former vice president of research at OpenAI, left the company in 2020 due to disagreements over OpenAI's direction in security. Since then, Anthropic has also faced its own challenges, with notable departures such as Mrinank Sharma, who warned about the pressures on AI companies to neglect security. Despite these departures, the security leadership at Anthropic has remained relatively stable compared to OpenAI.

Andrew Curran, an AI analyst, compared OpenAI's situation to a "curse," referencing a famous literary saga. This comparison underscores the seriousness of the situation and the potential impact of these departures on OpenAI's ability to maintain secure AI development.

A History of Notable Departures

Several prominent figures in security at OpenAI have left the company in recent years. Among them, Ilya Sutskever, co-founder and chief scientist, departed in May 2024 to establish Safe Superintelligence, a lab dedicated to AI safety. Jan Leike, who followed shortly after, joined Anthropic to continue his alignment research. These departures were followed by those of Miles Brundage, Lilian Weng, Andrea Vallone, Aleksander Madry, and Joshua Achiam, each expressing concerns or personal reflections about their departure.

Miles Brundage, senior advisor for AGI preparedness, left OpenAI in October 2024, voicing doubts about the company's readiness for the potential impacts of AI. Lilian Weng, former vice president of research and security, departed in November 2024, praising the work environment at OpenAI nonetheless. Andrea Vallone, head of model policy security research, left at the end of 2025 without publicly criticizing OpenAI.

Aleksander Madry, former head of preparedness, left in May 2026 after being moved from his leadership role. Joshua Achiam, former chief futurist, departed in July, sharing personal reflections on his departure without providing a specific reason. Finally, Johannes Heidecke, after leading security systems since 2021, decided to leave as OpenAI integrates security work more closely with its research.

These mass departures raise crucial questions about OpenAI's ability to maintain secure AI development, a central issue for the future of the company and the industry as a whole.