OpenAI Shuffles GPT-3 Research Team for Diversity, Not Performance, in Wake of Crackdown on GPT-3 Chat Bots

September 9, 2025

OpenAI has announced a major reorganization of the team that crafts ChatGPT’s personality, a change that reflects the company’s efforts to improve the way its AI communicates with people. The overhaul also comes at a time when artificial intelligence systems like ChatGPT are playing an increasing role in education, customer service, mental health support, and other forms of everyday digital communication.

Shaping the AI Personality

Since its launch, ChatGPT has received continuous upgrades aimed at enabling more natural and contextually consistent conversations. Early iterations were praised for their knowledge and responsiveness but were sometimes criticized for being cold or lacking emotional nuance.

With the release of GPT-5, users noted that while the model was extremely accurate and intelligent, its messages occasionally felt more neutral or civil than previous versions.

To address this, OpenAI began training GPT-5 to be “friendlier and warmer”, while maintaining accuracy and ethical safeguards. These refinements underscored a critical insight: intelligence alone is insufficient—AI must also communicate in ways that feel human and empathetic.

What the Model Behavior Team Does

The Model Behavior team, comprising approximately 14 researchers, has been instrumental in these advancements. Their focus includes:

• Reducing sycophancy, where AI models excessively agree with users
• Detecting and mitigating political and cultural biases
• Ensuring responses are accurate, balanced, and contextually sensitive

Joanne Jang, founding leader of the Model Behavior team, has been pivotal in shaping these advancements. Under her leadership, the team:

• Experimented with various conversational styles
• Studied user interactions
• Created guidelines for AI to address sensitive topics responsibly, including handling prompts from users expressing harm, distress, or emotional vulnerability

This work is especially crucial given ChatGPT’s widespread use in personal and emotional contexts.

Integration with Core Model Development

In August 2025, OpenAI announced that the Model Behavior team would merge with the Post Training group, led by Max Schwarzer. This move aims to embed personality development into the core model training process, rather than treating it as an afterthought.

Key objectives of this integration include:

• Making AI models inherently more empathetic and context-aware
• Addressing potential issues early in development, from tone nuances to ethically sensitive topics
• Enhancing ChatGPT’s ability to balance accuracy, safety, and human-like responses

Addressing Real-World Challenges

The reorganization comes amid increased scrutiny of AI behavior. High-profile incidents, such as a teenager expressing suicidal feelings to ChatGPT, highlighted the AI’s limitations in providing adequate emotional support.

OpenAI has responded with additional safeguards, including:

• Improved detection of distress signals
• Stronger guidance on sensitive topics
• Expanded user control features

These steps aim to ensure ChatGPT remains a safe, ethical, and supportive tool, particularly for vulnerable users seeking guidance or comfort.

The Future of AI Interaction

Incorporating personality research into core model development reflects a broader trend in the AI field. Experts, like Justine Cassell from Carnegie Mellon University, emphasize that how AI speaks is as important as what it speaks.

Users increasingly expect AI systems to be:

• Emotionally intelligent
• Sensitive to nuanced cues
• Able to respond in ways that feel human and empathetic

By merging behavioral knowledge with technical development, OpenAI ensures that AI personality is central to design, creating systems that are reliable, empathetic, and ethically responsible.

Joanne Jang’s new role as head of OAI Labs underscores this forward-looking approach. The lab focuses on:

• Exploring innovative ways for humans and AI to collaborate
• Prototyping interfaces that go beyond conversational interactions
• Creating dynamic, interactive AI experiences

This indicates that OpenAI is not only refining AI dialogue but reimagining the broader ways AI integrates into daily life.

Balancing Innovation with Responsibility

OpenAI’s reorganization sends a clear message:

• Developing state-of-the-art models is only part of the challenge
• Ensuring models act responsibly, communicate ethically, and provide meaningful interactions is equally critical

By integrating personality research into foundational development, OpenAI acknowledges that AI must be designed with the full spectrum of human communication in mind.

As AI technology progresses, OpenAI’s strategy demonstrates that it is possible to advance technical capabilities while maintaining trust, ethical standards, and emotional intelligence. This approach lays the groundwork for AI systems that are not just powerful tools, but thoughtful companions in human digital life.

Conclusion

Reframing the Model Behavior team represents a major strategic shift in OpenAI’s development philosophy. With the integration of behavioral research into core model training and the establishment of OAI Labs, OpenAI is moving toward AI that is both technically sophisticated and emotionally intelligent.

For ChatGPT users, this means that future interactions will likely feel more natural, empathetic, and contextually aware. For the AI industry, it highlights the importance of innovating responsibly, reinforcing that AI should be judged not only by its intelligence but also by the quality of human interaction it delivers.