OpenAI reorganizes research team behind ChatGPT’s personality

OpenAi reorganizes the model behavior team, a small but influential group of researchers who form how AI models interact for the company with people, according to Techcrunch.

In the August Memorandum of employees who saw TECHRUNCH, the chief research official at Openai Mark Chen said that the typical behavior team-which consists of about 14 researchers-joined the later training team, a larger research group responsible for improving the company’s artificial intelligence models after initial training.

As part of the changes, the model behavior team will now submit reports to Max Schwarzer in Openai. Openai spokesman confirmed these changes on Techcrunch.

Joan Gang, the founding leader of the model behavior team, Joan Gang, is moving to the start of a new project in the company. In an interview with Techcrunch, Jang says she is building a new research team called Oai Labs, which will be responsible for “inventing new facades of how people cooperate with artificial intelligence.”

The typical behavior team has become one of the main research groups in Openai, who is responsible for forming the character of the company’s artificial intelligence models alone from Sycophance – which happens when artificial intelligence models agree with the user’s beliefs and promotion, and even those unhealthy, rather than providing balanced responses. The team also worked to move in political bias in model responses and helped Openai in determining its position on artificial intelligence.

In the employee note, Chen said the time has come now to make the Openai model behavior team to develop the basic model. By doing this, the company indicates that its artificial intelligence “personality” is now a crucial factor in how technology evolves.

In recent months, Openai has faced the increasing scrutiny about the behavior of artificial intelligence models. Users strongly objected to the personal changes made on the GPT-5, which the company said has shown rates lower than SYCOPHANCY but it looked more cold for some users. This has restored access to some old models, such as GPT-4O, and the release of update to make the latest GPT-5 responses feel “warmer and more clear” without increasing the calendar.

Openai and all AI developers should walk for a thin line to make AI Chatbots by chance to speak with them but not Sycophantic. In August, the 16 -year -old boy filed a lawsuit against Openai for the alleged role of Chatgpt in their son’s suicide. The boy, Adam Rin, killed some of his ideas and committed suicide for chat plans (specifically a copy supported by GPT-4O), according to court documents, in the months before his death. The lawsuit claims that GPT-4O failed to retract his suicide thinking.

The typical behavior team has worked on each OpenAI model since GPT-4, including GPT-4O, GPT-4.5 and GPT-5. Before starting the unit, Jang previously worked on projects like Dall-E2, early Openai tool for photo process.

Jang announced in a post at X last week that it is leaving the team “to start something new in Openai”. He was the former head of the model with Openai for nearly four years.

Jang Techcrunch has told that she would work as Oaii Labs’s general manager, which will present to Chen at the present time. However, it is in the early days, and it is not clear after these new facades are.

“I am really excited to explore patterns that go beyond the chat model, which is currently more associated with companionship, or even agents, as there is a focus on autonomy,” Gang said. “I was thinking about [AI systems] As tools for thinking, making, playing, doing, learning, and calling. “

When asked if Oai Labs would cooperate in these new facades with the former Apple Design Head of Jony Ivee – who is now working with Openai on a family of AI – Jang said it is open to many ideas. However, she said she would likely start with the fields of research she knows more.

This story was updated to include a link to Jang’s Post, which announced its new website, which was released after the publication of this story. We also explain the models on the Openai Model behavior team.