OpenAI Implements Measures for Teen Safety on ChatGPT

OpenAI’s chief executive Sam Altman has announced that the company is enhancing safeguards for teenagers engaging with ChatGPT, especially as the US Senate conducts a hearing on the mental health implications of AI companions.

In a blog post on Tuesday, Altman recognised the challenges involved. He explained that it is essential to differentiate users who are under 18 from those who are not. OpenAI is currently developing a system that predicts age based on usage patterns. He stated that in cases of uncertainty, the default will be to provide the under-18 experience. In certain instances or regions, they may also require identification.

New Restrictions for Teen Users

Altman mentioned that the platform would implement restrictions on chatbot interactions with teenagers, specifically avoiding flirtatious conversations and discussions surrounding suicide and self-harm, even when presented in a creative context. He emphasised that if a user under 18 expresses suicidal thoughts, the company will make efforts to reach out to the user’s parents, and if unsuccessful, they will notify authorities in the case of immediate danger.

Parental Controls and Its Background

Earlier this month, OpenAI revealed plans for new parental control features that will enable account linking to parents, the ability to disable chat history, and alerts if ChatGPT identifies a teenager in distress. These measures are in response to a lawsuit filed by the family of Adam Raine, a teenager who tragically died by suicide following prolonged interactions with ChatGPT.

Emotional Testimony from Adam’s Father

During the hearing, Adam’s father, Matthew Raine, delivered a profound statement: “ChatGPT spent months coaching him toward suicide. As parents, one cannot fathom what it’s like to read a conversation with a chatbot that groomed your child to take his own life. What started as a homework aid transformed into a confidant and ultimately a suicide coach.”

Raine revealed that the chatbot made references to suicide 1,275 times in talks with his son. He urged Altman to withdraw GPT-4o from public access until sufficient safety measures are established.

Public Health Concerns

Another parent, who spoke under the alias Jane Doe, addressed lawmakers: “This is a public health crisis. We are in a mental health battle, and it truly feels like we are losing.”