4 Minutes
Some people treat ChatGPT like a search box. Others talk to it the way they would text a friend at 2 a.m. when sleep will not come. That second use case has pushed AI into territory far heavier than product demos and coding help, and OpenAI is now responding with a feature designed for moments when a conversation appears to turn dangerous.
The new tool, called Trusted Contact, is beginning to roll out in ChatGPT settings for adult users. Its purpose is simple on paper: let a user choose one person who can be notified if the system detects a serious risk of self-harm. In practice, OpenAI is trying to build a safety net without turning private chats into automatic alerts.
Setting it up is voluntary. Users can nominate one trusted person, but that person must meet the age requirement: at least 18 years old in most regions, or 19 in South Korea. Once selected, the contact receives an invitation explaining the role and has seven days to accept. If they decline, the user can choose someone else.
What happens next is where OpenAI is clearly trying to avoid overreach. Alerts are not sent automatically the moment a system detects concerning language. First, ChatGPT tells the user that their Trusted Contact may be notified. It also encourages the user to reach out directly, even offering suggested ways to start that conversation. After that, a small group of specially trained human reviewers examines the case. Only if they confirm that the situation points to a serious self-harm risk does OpenAI send a notification.
That notification can arrive by email, text message, or through the app. It does not include chat logs, quotes, or a summary of the conversation. Instead, it simply tells the trusted person that self-harm was mentioned in a concerning context and asks them to check in. OpenAI says it aims to complete the human review in less than an hour.
Where this fits in OpenAI’s broader safety push
This is not appearing in isolation. OpenAI has already introduced safety features tied to teen accounts, including alerts for parents when linked accounts show signs of distress. Trusted Contact extends that idea to adults, which says a lot about how the company now sees ChatGPT being used. The chatbot is no longer just a productivity assistant. For many users, it has become a place for deeply personal, emotionally loaded conversations.
According to OpenAI, the feature was developed with input from clinicians, researchers, and mental health groups, including the American Psychological Association. That outside involvement matters. Any tool that touches self-harm risk, privacy, and crisis response sits in a sensitive space where design decisions carry real consequences.
There are also clear limits. Trusted Contact is not a substitute for a crisis hotline, emergency services, or professional mental health support. OpenAI says ChatGPT will continue directing users to those resources when needed. Users can remove or change their chosen contact at any time, and the contact can opt out whenever they want.
There is a quieter message inside this rollout. OpenAI is acknowledging something the tech industry has danced around for a while: people are already bringing their most vulnerable moments to AI chatbots. Whether companies intended that or not no longer matters much. The behavior is here. Features like Trusted Contact are an attempt to respond to that reality with a little more caution, a little more humanity, and a recognition that even the most advanced chatbot should not be the only voice in the room when things get dark.
Leave a Comment