OpenAI taps psychology experts for new mental well-being council

When millions of teenagers start treating an AI chatbot like a personal diary or therapist, standard content filters simply don't cut it anymore.

OpenAI just launched its new Expert Council on Well-Being and AI, an advisory group tasked with shaping how ChatGPT and Sora interact with humans. The goal is to define what healthy, supportive AI usage actually looks like across different age groups, with a heavy emphasis on mental health and emotional well-being.

Heavyweight researchers meet the product team

Rather than relying strictly on in-house safety engineers, OpenAI recruited six external researchers across psychology, psychiatry, pediatrics, and human-computer interaction.

The council includes David Bickham from Boston Children’s Hospital and Harvard Medical School, Mathilde Cerioli of everyone.AI, and Munmun De Choudhury from Georgia Tech. They are joined by Northwestern University’s David Mohr, Hunter College professor and Arcade Therapeutics co-founder Tracy Dennis-Tiwary, and public health veteran Robert K. Ross.

The group kicked off last week with an in-person workshop at OpenAI, sitting down directly with the teams building the core models and consumer features.

Fine-tuning empathy and crisis alerts

This collaboration is not starting from zero. OpenAI noted that it had already been consulting several of these specialists informally while developing its parental controls.

One clear example is the notification sent to parents when a teen appears to be in emotional distress. The council members helped rewrite that alert copy, aiming for a supportive tone that avoids sounding punitive to the teenager while still giving parents clear, actionable guidance.

Alongside this council, OpenAI is also leaning on a separate Global Physician Network. That group brings in practicing clinicians across psychiatry, pediatrics, and crisis intervention to stress-test how ChatGPT responds to delicate real-world prompts.

Still, OpenAI made sure to include a clear caveat: while the council provides guidance and guardrails, the company itself retains full ownership of every final decision.

Why it matters

For everyday users, particularly parents and students, AI is increasingly blurring into emotional territory. Whether people should use chatbots for emotional support is almost beside the point now, because they already are. Having pediatric and mental health experts help set interaction guardrails makes ChatGPT noticeably safer around vulnerable teens.

For the AI industry, OpenAI is clearly learning from social media's past mistakes. Big Tech spent the last decade getting grilled by regulators over teen depression and screen addiction. By establishing an external council filled with Harvard and hospital-affiliated researchers before major lawsuits land, OpenAI is playing proactive defense.

We think involving clinicians in tone and prompt design is a genuinely smart move that beats clumsy keyword blocking. But there is an obvious blind spot here. The council is strictly advisory, and OpenAI openly stated that final calls remain internal. When engagement metrics inevitably clash with digital well-being recommendations, whether OpenAI actually listens to these experts will be the real test.