OpenAI has recently reorganised the research team responsible for shaping ChatGPT’s personality, which has significant implications for how users experience the AI.
The Model Behaviour team, consisting of about 14 researchers, has been merged into the larger Post Training team led by Max Schwarzer.
This change seeks to integrate AI personality into basic model development. Chief Research Officer Mark Chen announced the restructuring in a message to match personality work with model refinement, signalling a strategic shift in OpenAI’s priorities.
Model behaviour team’s new role and leadership change
The Model Behaviour team has been pivotal in defining how ChatGPT interacts with users, focusing on balancing friendliness and avoiding excessive flattery or sycophancy. Despite their efforts, GPT-5’s launch faced criticism for featuring a colder, less engaging tone.
This led OpenAI to restore access to previous versions like GPT-4o and to update GPT-5’s responses to feel “warmer and friendlier.” The reorganisation directly responds to these user critiques and legal scrutiny, such as a lawsuit alleging ChatGPT failed to address a user’s suicidal thoughts adequately.
Meanwhile, the former team leader, Joanne Jang, is stepping away to start OAI Labs, a new initiative to explore novel AI-human collaboration interfaces.
Balancing AI personality with ethical interaction
OpenAI’s integration of the Model Behaviour team into Post Training shows a shift from treating AI personality as a separate feature to making it a core aspect of AI development.
The company aims to improve ChatGPT’s response by blending warmth with honesty and safety. This change reflects OpenAI’s commitment to refining user experience while addressing ethical concerns like political bias and harmful sycophancy.
OpenAI is working to perfect a user-friendly interaction style that maintains robustness and reduces risks, especially in sensitive cases.
Joanne Jang’s new focus at OAI Labs may also drive breakthroughs in AI interfaces beyond traditional chatbots, emphasising collaboration in AI-human relationships.
This reorganisation marks a defining shift in how OpenAI shapes ChatGPT’s personality. It is closely tied to the AI’s foundational training and post-training refinements to meet evolving user expectations and ethical standards.