OpenAI, the company behind the popular ChatGPT AI model, recently shook up one of its most important research teams. This move happened as the tech giant faced growing questions about how its AI models act and whether they can be trusted. The company decided to merge its “Model Behavior” team, which sets how AI models respond, with the “Post Training” group.
This change is more than just an internal reshuffle. It comes at a time when the “personality” of AI models and how they interact with people has become a huge deal. Users want AI that is not only smart but also friendly and fair. There’s also increasing pressure on AI companies to address concerns about bias and how their systems handle sensitive topics. Cryptopolitan first reported this significant internal shift.
Defining AI’s Personality
The Model Behavior team, made up of about 14 researchers, played a big part in shaping the “personality” of OpenAI’s models. Their work focused on making sure the AI didn’t just agree with users blindly, even on harmful ideas. This tendency is called “sycophancy.” The team also worked to fix political bias in the AI’s answers.
Mak Chen, OpenAI’s research director, explained the reasons for the merger in a memo. He said combining Model Behavior with Post Training will bring the work on personality and safety closer to the main development of the models. This means improvements in how AI acts and its safety features will be built into the technical process more directly. Chen believes this is the perfect time for the teams to work together, as AI models are learning to interact with users in much more complex ways.
Joanne Jang’s New Vision
With the reorganization, there’s also a new direction for the team’s former leader. Joanne Jang, who founded and led the Model Behavior team, is now starting a new research lab within OpenAI called OAI Labs.
Jang recently shared that her new group aims to “invent and prototype new interfaces for how people collaborate with AI.” She wants to move past simple chat interactions. Her goal is to explore new ways for humans and AI to work together, focusing on sparking creativity, giving users more control, and promoting shared learning. Jang noted that the initial work will be research-focused. However, she hinted at possible future collaborations with hardware projects backed by OpenAI, perhaps even with Jony Ive, Apple’s former design chief.
Facing Public Scrutiny
This internal shake-up comes as OpenAI deals with a wave of criticism over how its latest models behave. Users pointed out that GPT-5, while less overly agreeable, seemed to have a “colder” personality. This feedback led OpenAI to bring back access to older versions like GPT-4o. They also released updates to strike a better balance between being helpful and being accurate.
The discussion about AI companies’ responsibilities grew much more serious after a lawsuit. The parents of Adam Raine, a 16-year-old, sued OpenAI, claiming negligence. Reports say Adam shared suicidal thoughts with ChatGPT (a version using GPT-4o) before his death. The system allegedly did not respond to these thoughts in a proper way. This heartbreaking case highlights how important the work of teams like Model Behavior is. It stresses that AI interactions must be not just useful, but also safe and ethical.
The Road Ahead for OpenAI Research
The future for OAI Labs isn’t fully set yet. Still, Jang is excited to find new ways for people to interact with AI. She hopes to move beyond the idea of AI simply being a “companion,” which is often how current chatbots are seen. She believes AI systems should be tools to “connect, learn, and think,” not just stand-alone agents.
For now, Mak Chen will directly oversee OAI Labs. Creating this new lab shows OpenAI’s plan to explore many different ways humans and machines can interact. The company is working to keep its users’ trust. This is vital in an industry that faces more and more questions about the social and ethical dangers of artificial intelligence.