OpenAI Restructures Model Behavior Team; Joan Zhang Launches New Project OAI Labs

Staff Reporter | PNN

OpenAI is restructuring its Model Behavior team. The team, consisting of around fourteen researchers, will now operate under OpenAI’s Post-Training group. As part of this change, the team will report to Max Schwarzer, head of Post-Training.

The team’s founder, Joan Zhang, is launching a new project called OAI Labs. She stated that the new team will “invent and prototype new interfaces for human-AI collaboration.”

The Model Behavior team has been working on shaping the personality of OpenAI’s AI models, reducing political bias, and minimizing compliance-driven responses (psychofancy). The team has contributed to models including GPT-4, GPT-4.5, and GPT-5.

OpenAI’s Chief Scientist, Mark Chen, said that the work of the Model Behavior team is now more closely integrated with core model development. This highlights that AI “personality” has become a critical aspect of technological progress.

Through the restructuring of the research team and the launch of OAI Labs, OpenAI aims to take AI interaction and collaboration with users to a new level. Joan Zhang emphasized that she envisions AI systems as tools for thinking, learning, creating, and connecting—beyond the limits of chat or agent functionality.