OpenAI is reshaping its Model Behavior team, a small but influential research group tasked with shaping how the company’s AI models interact with users.
Internal Reorganization
In an August memo to staff, Chief Research Officer Mark Chen said the 14-person Model Behavior team will merge with the Post Training team, which oversees refining models after their initial pre-training. From now on, the group will report to Post Training lead Max Schwarzer.
The team’s founding leader, Joanne Jang, is moving on to start a new initiative within the company. She will head a new research group called OAI Labs, which aims to design and prototype new ways people can collaborate with AI.
The Model Behavior team has been central to shaping how OpenAI models present themselves, reducing sycophancy (when AI agrees with users without challenge) and navigating politically sensitive responses. The group also helped define OpenAI’s position on whether AI should be seen as conscious.
OpenAI says the reorganization brings this work closer to core model development, signaling that model “personality” is becoming as critical as raw technical performance.
User Reactions and Challenges
The company has faced increasing criticism over how its models interact with users. Some GPT-5 users objected to personality shifts that made the system feel colder, even though sycophancy rates were reduced. This backlash led OpenAI to restore access to legacy models like GPT-4o and adjust GPT-5 to feel “warmer and friendlier.”
The issue of AI behavior also carries high stakes. In August, the parents of 16-year-old Adam Raine sued OpenAI, alleging ChatGPT failed to challenge his suicidal ideations before his death. The case has amplified scrutiny of how AI models should respond in sensitive situations.
Jang, who previously contributed to projects like Dall-E 2, has led the Model Behavior team through the development of GPT-4, GPT-4o, GPT-4.5, and GPT-5. She confirmed in a post on X that she is leaving the team to focus on OAI Labs, which will report to Chen in the near term.
“I’m excited to explore patterns that move us beyond the chat paradigm,” Jang said, describing AI as tools for “thinking, making, playing, doing, learning, and connecting.” She added she is open to a wide range of research directions, though initial efforts will likely draw from areas she knows best.
What The Author Thinks
OpenAI’s decision to fold the Model Behavior team into its core development pipeline shows that AI personality is no longer an afterthought. As users increasingly rely on chatbots for sensitive conversations, an AI’s tone, empathy, and ability to push back matter as much as accuracy. If personality is neglected, the result could be sterile machines that alienate people—or worse, systems that reinforce harmful thinking. In the long run, the winners in AI will be those who balance raw intelligence with humanity.
Featured image credit: Solen Feyissa via Unsplash
For more stories like it, click the +Follow button at the top of this page to follow us.