OpenAI reorganizes research team behind ChatGPT’s personality (techcrunch.com)

OpenAI has reorganized its influential Model Behavior team, which shapes the personality and interaction style of its AI models like ChatGPT, by merging it with the larger Post Training research group. This shift, confirmed by OpenAI and announced by chief research officer Mark Chen, signifies a strategic move to integrate the development of AI “personality” more closely with core model improvements. The roughly 14-member team, known for tackling challenges such as reducing sycophancy—where models overly agree with users—and managing political bias, will now report to Post Training lead Max Schwarzer. The reorganization highlights how crucial model behavior is to OpenAI’s evolving AI systems amid debates over maintaining user-friendly yet balanced interactions. Additionally, Model Behavior's founding leader Joanne Jang is stepping down to start a new internal research group called OAI Labs, aimed at pioneering novel AI interfaces beyond the conventional chat paradigm. Jang envisions these interfaces as tools for thinking, creating, and connecting, rather than purely conversational agents. Her new team may collaborate with OpenAI’s hardware initiatives but will initially focus on familiar research domains. This personnel and organizational shuffle comes amid user backlash over recent GPT-5 personality changes, lawsuits over ChatGPT’s handling of sensitive situations, and ongoing efforts to refine AI behavior to be both engaging and responsible, making this realignment a critical development for AI/ML research and deployment.