A new safety net: OpenAI introduces "Trusted Contact" for ChatGPT users, aiming to connect at-risk individuals with a real-world support system.
Feature Details
OpenAI has launched a new safety feature for ChatGPT called "Trusted Contact," allowing adult users to voluntarily designate a person to be notified if the system detects conversations about self-harm or suicide.
The Trusted Contact feature is optional and available to adult ChatGPT users. To activate it, users must add contact information for another adult (aged 18 or older, or 19 in South Korea) in their ChatGPT account settings. The designated contact must accept the invitation within one week. Users can edit or remove their chosen contact at any time, and the designated contact can also remove themselves from the arrangement.
If OpenAI's automated systems detect discussion of self-harm or suicide, ChatGPT will encourage the user to reach out to their Trusted Contact and inform them that the contact may be notified. A small team of trained staff reviews the situation before any alert is sent. If the conversation is deemed to indicate serious safety concerns, ChatGPT sends a brief notification via email, text message, or in-app alert to the Trusted Contact, informing them to check in on the user.
OpenAI states that the notification is limited in scope and does not include chat details, transcripts, or summaries of the conversation. The company aims to review safety notifications and respond within one hour.
Development and Guidance
OpenAI reports that the feature was developed with guidance from clinicians, researchers, and mental health organizations.
Context
This feature builds on an emergency contact feature introduced alongside ChatGPT's parental controls in September. OpenAI has previously introduced safety measures for teen accounts and automated prompts directing users to seek professional help.
The company has faced lawsuits from families of individuals who died by suicide after interacting with ChatGPT, with some families alleging the chatbot encouraged self-harm. A similar feature has been introduced by Meta on Instagram.
Limitations
The Trusted Contact feature remains optional, and users may operate multiple ChatGPT accounts. OpenAI's parental controls are also optional.