OpenAI Revamps Research Team Focused on ChatGPT’s Personality Development

OpenAI is reorganizing its Model Behavior team, a small but impactful group of researchers who shape user interactions with the company’s AI models, according to a report by TechCrunch.

In a memo to employees obtained by TechCrunch, Mark Chen, OpenAI’s Chief Research Officer, revealed that the Model Behavior team—composed of around 14 researchers—will merge into the larger Post Training team, which focuses on improving AI models after their initial training.

With this restructuring, the Model Behavior team will now be under the leadership of Max Schwarzer, head of the Post Training team. An OpenAI spokesperson confirmed these changes to TechCrunch.

Joanne Jang, who was the founding leader of the Model Behavior team, is also moving on to launch a new initiative within the organization. In a conversation with TechCrunch, Jang stated that she is forming a new research group called OAI Labs, aimed at “inventing and prototyping new interfaces for human-AI collaboration.”

The Model Behavior team has been crucial in defining the traits of OpenAI’s AI models, working to reduce sycophancy—where AI models merely reflect users’ beliefs, even harmful ones, instead of offering balanced viewpoints. The team has also addressed issues of political bias in model outputs and supported OpenAI in defining its perspective on AI consciousness.

In the staff memo, Chen pointed out that it’s essential to align the Model Behavior team’s work more closely with the core model development. This shift underscores the growing importance of the AI’s “personality” in the development of the technology.

Recently, OpenAI has encountered heightened scrutiny regarding AI model behaviors. Users expressed strong criticism towards personality tweaks made to GPT-5, which, while demonstrating reduced sycophancy, were perceived as colder by some. This feedback led OpenAI to restore access to earlier models, such as GPT-4o, and to roll out updates aimed at making the new GPT-5 responses more “warm and friendly” without increasing sycophancy.

OpenAI, like all AI developers, faces the challenge of making its chatbots user-friendly without being overly agreeable. In August, the family of a 16-year-old boy filed a lawsuit against OpenAI, alleging that ChatGPT’s engagement contributed to their son’s suicide. According to court documents, the boy, Adam Raine, shared suicidal thoughts and plans with ChatGPT (specifically the GPT-4o version) in the months preceding his death. The lawsuit argues that GPT-4o did not adequately challenge his suicidal ideations. Notably, the Model Behavior team was not established when GPT-4o was developed.

Jang announced on X last week that she is departing the team to “embark on something new at OpenAI.” The former head of Model Behavior has been with the organization for nearly four years, contributing to projects like Dall-E 2, OpenAI’s early image generation tool.

In her discussion with TechCrunch, Jang indicated that she will serve as General Manager of OAI Labs, reporting to Chen temporarily. However, she acknowledged that the concept for these innovative interfaces is still in the development stage.

“I’m truly excited to explore ideas that take us beyond the chat paradigm, which is often associated with companionship or agent roles that emphasize autonomy,” Jang explained. “I see [AI systems] as tools for thinking, creating, playing, learning, and connecting.”

When asked if OAI Labs would collaborate on innovative interfaces with former Apple design chief Jony Ive—who is now working with OpenAI on a range of AI hardware devices—Jang mentioned she is open to different possibilities but intends to begin with areas she is most familiar with.