ai
OpenAI Adds Trusted Contact Feature to ChatGPT for Self-Harm Queries
OpenAI on Thursday launched an optional Trusted Contact safeguard for adult ChatGPT users that notifies a designated friend, family member or caregiver if the system detects serious self-harm or suicide-related discussions. The feature triggers automated encouragement for the user to reach out and, after human review, sends a brief alert to the contact without sharing conversation details.
May 7, 5:53 PM·2m read3 sources
FI
gizmodo.comAudio version
Tap play to generate a narrated version.
OpenAI announced Thursday an optional safety feature called Trusted Contact for its ChatGPT service that allows adult users to designate another adult as an emergency contact for mental health concerns. The feature notifies the chosen contact if OpenAI's systems detect that a user may be discussing self-harm or suicide.
It aims to connect people in potential crisis with someone they know and trust. Users over 18 can enable the setting by adding contact details for another adult. The designated Trusted Contact must accept the invitation within one week. Either party can later remove the connection at any time.
When the system flags a conversation as potentially involving self-harm, ChatGPT encourages the user to reach out to their Trusted Contact. A small team of specially trained reviewers then assesses the situation. If reviewers determine there is a serious safety risk, an automated alert is sent to the contact via email, text message or in-app notification.
The message is intentionally brief and does not include chat transcripts or details to protect user privacy.
The new capability expands an emergency contact system that the company previously introduced for teenage accounts. It also complements existing automated responses that direct users toward professional helplines when conversations trend toward self-harm topics.
OpenAI currently combines automation and human review to handle harmful incidents. Conversational triggers alert the system to suicidal ideation, after which the information goes to a human safety team. The company states that every such notification receives human review, with a goal of completing reviews in under one hour.
The Trusted Contact rollout comes amid a wave of lawsuits from families who say relatives committed suicide after interacting with the chatbot. In some cases, families allege the service encouraged self-harm or assisted in planning it. The company has faced criticism over its previous safeguards.
