OpenAI is introducing a 'Trusted Contact' feature to ChatGPT. It is an optional safety feature that allows users to notify family, friends, or guardians they have designated in crisis situations.
This feature is being rolled out sequentially to adult users aged 18 and older. In Korea, it is available to users aged 19 and older. Users can register one trusted contact in the ChatGPT settings. The registered person must receive an invitation and accept it within one week for the feature to be activated.
Trusted Contact activates when a user displays serious risk signals related to self-harm or suicide. When ChatGPT's automation system detects a potential risk, it first informs the user that it may notify a Trusted Contact. Subsequently, a specially trained reviewer verifies the situation. If the reviewer determines there is a serious safety concern, an alert is sent to a Trusted Contact.
Notifications may be delivered via email, text message, ChatGPT app notifications, etc. However, conversation content or the entire chat history is not shared. The notification includes a general reason that the user may have engaged in a concerning conversation related to self-harm, and instructions to contact the person involved for verification.
OpenAI explained that this feature does not replace emergency services or professional treatment. ChatGPT continues to guide users to seek help from crisis hotlines, emergency services, mental health professionals, or people around them if necessary. Trusted Contacts are more of an auxiliary safety net designed to connect users with real-life relationships.
This feature is an extension of the existing guardian notification feature for youth accounts to adult users. OpenAI has been providing safety alerts within its parental control features to notify guardians when there are signs of acute distress in youth accounts, and now it allows adult users to designate a trusted person themselves.
OpenAI stated that it consulted with clinicians, researchers, and suicide prevention organizations while developing its trusted contact feature. Arthur Evans, CEO of the American Psychological Association, explained that social connection is an important protective factor during times of emotional distress. OpenAI also revealed that a global network of over 260 licensed physicians from 60 countries and an AI Wellbeing Expert Committee participated in the development process.
Users can change or delete trusted contacts at any time. People designated as trusted contacts can also decline invitations or withdraw if they do not wish to be included. According to OpenAI Help, this feature is available for personal ChatGPT accounts and does not apply to shared workspaces such as ChatGPT Business, Enterprise, or Edu.
Trusted Contact is an example of how AI services should handle sensitive conversations. Rather than chatbots attempting to solve every problem directly, the approach is to connect users with real people when warning signs appear. At the same time, a balance between automated detection, human review, and privacy has become crucial. OpenAI explained that all notifications undergo verification by trained reviewers before being sent.
