OpenAI is rolling out a new optional safety feature for ChatGPT, allowing adults to assign an emergency contact for mental health and safety issues. A designated “Trusted Contact,” which can be a friend, family member, or caregiver, will receive notifications if OpenAI detects discussions around self-harm or suicide between a user and the chatbot.
“Trusted Contact is built on a straightforward, expert-supported principle: reaching out to a familiar and trusted person during a crisis can significantly help,” OpenAI stated. “It provides an extra layer of support in addition to the localized helplines available within ChatGPT.”
The opt-in Trusted Contact feature is available to all adult ChatGPT users, who can enable it by inputting an adult’s contact details (18+ globally or 19+ in South Korea) into their ChatGPT account settings. The Trusted Contact must accept the invitation within a week. Users can remove or update their contact, and the Trusted Contact can also opt out anytime.
OpenAI clarifies that notifications are “intentionally limited,” and no chat details or transcripts are shared with the Trusted Contact. When OpenAI’s systems identify discussions about self-harm, ChatGPT urges the user to contact their Trusted Contact for assistance and informs them that the contact might be notified. A “small team of specially trained people” will review the situation, and ChatGPT will notify the Trusted Contact through email, text, or in-app notification if the conversation indicates significant safety concerns.
This feature expands upon the emergency contact option introduced alongside ChatGPT’s parental controls in September, following a tragic case where a 16-year-old died by suicide after confiding in ChatGPT for several months. Similarly, Meta introduced a feature involving parent alerts if their children “repeatedly” search for self-harm topics on Instagram.