OpenAI Trusted Contact: How ChatGPT’s Opt-In Self-Harm Alert Works

OpenAI’s Trusted Contact is designed to make ChatGPT do something a chatbot cannot do alone: involve a real person the user has chosen in advance. The opt-in feature lets adult ChatGPT users name a trusted contact, such as a friend or family member, who may be alerted if a conversation suggests self-harm or suicide risk .

What Trusted Contact does

Trusted Contact is a safety setting for adult ChatGPT users. After a user opts in and designates someone, ChatGPT may encourage the user to reach out to that person if self-harm ideation appears in a conversation .

The feature can also send the chosen contact an automated alert encouraging them to check in with the user . That makes it different from a standard crisis-resource message: it can add a human support layer involving someone the user selected ahead of time .

How the alert flow works

The described flow has three parts:

1. Opt in first. The adult user chooses a trusted person in advance, rather than the feature being automatically assigned to everyone .
2. A risky conversation is detected. The safeguard is tied to possible self-harm or suicide ideation in a ChatGPT conversation .
3. A check-in can be triggered. ChatGPT can prompt the user to contact their trusted person and can send that person an automated alert asking them to check in .

Public rollout reporting describes the alert as a check-in prompt. It does not establish that the trusted contact becomes a crisis responder, clinician, or emergency service .

Why OpenAI is adding this layer

Trusted Contact fits into OpenAI’s broader effort to strengthen ChatGPT’s responses in sensitive mental-health conversations. OpenAI says its safeguards include training models not to provide self-harm instructions, shifting toward supportive and empathetic language, and steering people toward crisis resources when appropriate .

OpenAI has also said it worked with more than 170 mental-health experts to improve how ChatGPT recognizes distress, responds with care, and guides people toward real-world support . In that context, Trusted Contact moves beyond the chatbot’s reply itself by creating a path for a user-selected person to notice and check in .

What users should understand before enabling it

The most important practical detail is consent: Trusted Contact is opt-in, and the user has to choose the contact first . That means the feature depends on selecting someone who would be appropriate and able to respond if they receive a check-in alert.

The feature should also be understood as a safety prompt, not a substitute for emergency help. OpenAI’s own safety materials frame ChatGPT’s role as recognizing distress, responding carefully, and guiding people toward real-world support and crisis resources . If someone is in immediate danger, local emergency or crisis resources remain the appropriate route rather than relying on a chatbot notification.

What is still unclear

The sourced rollout explains the basic mechanism: an adult user opts in, ChatGPT detects possible self-harm or suicide-risk signals, and an automated alert can ask the trusted contact to check in . It does not provide independent outcome data showing whether Trusted Contact reduces harm in real-world use.

It also describes the contact receiving an automated check-in alert, but the available reporting cited here does not specify the full contents of that alert or whether any conversation excerpts are shared . Users considering the setting should review the current ChatGPT product flow and privacy details before enabling it.

Bottom line

Trusted Contact is OpenAI’s new opt-in safety feature for adult ChatGPT users who want a preselected person to be reachable during possible self-harm or suicide-risk conversations. If enabled, ChatGPT can encourage the user to reach out and can alert the chosen person to check in, adding a human support layer to OpenAI’s existing crisis-response safeguards .