TechCrunch learned that OpenAI was reorganizing its model of model behavior, a tiny but influential group of researchers who shape the interaction of the company’s AI models.
In the August note to the staff seen by TechCrunch, director of OPENAI research, Mark Chen, said that the model of model behavior-which consists of about 14 researchers-pads to the Post Training team, a larger research group responsible for improving the company’s AI models after their initial initial training.
As part of the changes, the behavior team The model will now report to the Openai after training Max Schwarzer. Openai spokesman confirmed these changes in TechCrunch.
The leader of the Behavior Model Founder, Joanne Jang, is also starting a fresh project in the company. In an interview with TechCrunch, Jang says that he is building a fresh research team called OAI Labs, which will be responsible for “inventing and prototyping new interfaces for how people work with AI.”
Model behavior team has become one of the key research groups of OPENAI, responsible for shaping the personality of the company’s AI models and reducing rewarding – which occurs when AI models simply agree and strengthen users’ beliefs, even unhealthy, instead of offering balanced answers. The team also worked on moving after political warning in models and helped OpenAI to define their position AI awareness.
In a note for Chen staff, he said that now it’s time to bring the work of the OpenAI model behavior team to the basic development of models. In this way, the company signals that the “personality” of its artificial intelligence is currently considered a key factor evolving technology.
In recent months, Opeli has met with greater analysis of the behavior of AI models. Users definitely opposed personality changes introduced to the GPT-5, which according to the company was shown by lower tilt indicators, but they seemed cooler for some users. This led OpenAI to restore access to some of its older models, such as GPT-4O, and the release of updates so that newer GPT-5 answers seem “warmer and more friendly” without increasing absorption.
TechCrunch event
San Francisco
|.
October 27-29 2025
Opeli and all AI models programmers have to go through a slim line so that their AI chatbots are amiable to conversation, but not sycophantic. In August, parents of a 16-year-old boy sued OpenAi over the alleged role of chatgpt in their son’s suicide. The boy, Adam Raine, confided in some of his suicidal thoughts and chatgpt plans (in particular the version powered by GPT-4O), in accordance with court documents, in the months preceding his death. The claim claims that the GPT-4O has not pushed its suicidal thoughts.
Model behavior team has worked on every OPENAI model since GPT-4, including GPT-4O, GPT-4.5 and GPT-5. Before starting the device, Jang has previously worked on projects such as Dall-E 2, a tool for generating an early OpenAI image.
Jang announced in Post on X Last week, he leaves the band to “start something new at Openai.” The former head of the behavior model has been at Openai for almost four years, and previously worked on projects such as Dall-E 2, a tool for generating an early OpenAI image.
Jang told Techcrunch that he would be the general director of OAI Labs, who will report to Chen for now. However, there are early days and it is not yet clear what these fresh interfaces will be, she said.
“I am very excited about the study of patterns that come out of the chat paradigm, which is currently more associated with the company and even agents in which he emphasizes autonomy,” said Jang. “I thought about [AI systems] As instruments for thinking, creating, playing, performing, learning and combining. “
Asked if Oai Labs would work with these pioneering interfaces with the former head of Apple Design Jons Ive – who is Now he works With OpenAI in the family of AI equipment – Jang said he was open to many ideas. However, Jang said that she would probably start with research areas that is more familiar.