OpenAI is restructuring its Model Behavior team, a group of 14 researchers responsible for shaping how its AI models interact with users. According to an internal memo from Chief Research Officer Mark Chen, the team will now merge with the larger Post Training division, which focuses on refining AI models after initial training. Moving forward, the group will report to Post Training lead Max Schwarzer, a move the company says is meant to align model behavior more closely with core development.
The reorganization comes as the founding leader of the Model Behavior team, Joanne Jang, transitions to a new role at OpenAI. She will lead OAI Labs, a new research initiative tasked with inventing and prototyping fresh ways for people to collaborate with AI. Jang, who has worked on every OpenAI model since GPT-4 and previously contributed to Dall-E 2, described her new project as an opportunity to explore interfaces that go beyond the traditional chat format.
Since its creation, the Model Behavior team has been central to shaping the “personality” of OpenAI’s models. Its work has focused on curbing sycophancy, reducing political bias, and guiding how the company approaches questions of AI consciousness. The changes come at a time when OpenAI is under scrutiny from users over adjustments to GPT-5’s personality and facing a lawsuit alleging ChatGPT failed to adequately respond to a teenager’s suicidal thoughts.
While Jang’s new unit will initially report to Chen, the company has not yet revealed the full scope of OAI Labs’ projects. In an interview, she hinted at experimenting with new patterns for AI use that emphasize creativity, learning, and connection. She also suggested openness to collaborations, though she indicated her immediate focus would remain within her research expertise, even as OpenAI explores partnerships with high-profile figures like former Apple design chief Jony Ive.