OpenAI is expanding ChatGPT's safety tools by adding a "trusted contact" feature for users aged 18 and over. The system's automated processes and human moderators can notify that designated contact if they determine the user is engaging in conversations suggestive of self-harm or otherwise at risk.
Why this was introduced
According to data previously shared by OpenAI, a small but non-negligible share of ChatGPT users show weekly signs of mental-health crises: about 0.07 percent exhibited indicators associated with psychosis or mania, 0.15 percent showed signs of self-harm or suicide risk, and another 0.15 percent displayed signs of excessive emotional attachment to the AI. The trusted contact feature is part of the company's response to these patterns.
How the feature works
- The option is available in settings only to users who are at least 18 years old.
- A user can designate another adult as a trusted contact; the named person receives an invitation by email and phone number.
- The invited contact has one week to accept or decline; if declined, the user may choose another contact.
- If the system or a human moderator sends a notification, the trusted contact is alerted via email, SMS, or an in-app message.
- Notifications are intentionally general: they state that self-harm concerns have arisen, encourage the contact to reach out to the user, and provide guidance on how to approach the conversation. They do not share detailed content of the user's conversations.
Connection to existing safety measures and expert involvement
The new feature builds on ChatGPT's earlier parental safety alerts, which notify parents when a linked teen account shows signs of trouble. OpenAI says it developed the trusted contact capability with input from mental-health and suicide-prevention professionals, as well as researchers and organizations.
What this is not
OpenAI stresses that the trusted contact notification is not a replacement for professional care or crisis hotlines. The company positions the feature as one layer in a multi-tiered safety system intended to support users in distress.
Summary
The trusted contact option gives adult ChatGPT users a way to name a trusted person who can be notified if the system or moderators detect self-harm risk. The feature sends limited, general alerts and was developed with expert input; OpenAI emphasizes it complements, rather than substitutes for, clinical or crisis services.
