Introducing Trusted Contact in ChatGPT
Key Points
- Trusted Contact notifies designated adults of serious self-harm concerns detected in ChatGPT conversations
- Feature developed with clinicians and mental health experts; human review required before notifications sent
- Privacy-first design shares only general concern reason, no chat transcripts or conversation details
Summary
OpenAI has launched Trusted Contact, an optional safety feature in ChatGPT that enables adults to designate a trusted person who may receive notifications if automated systems and trained reviewers detect serious self-harm concerns in conversations.
Key Points
- Feature Setup: Users 18+ (19+ in South Korea) can nominate one trusted adult from ChatGPT settings; the nominee must accept within one week to activate the feature
- Detection & Notification: Automated monitoring flags potential self-harm discussions; trained human reviewers validate concerns before sending notifications via email, SMS, or in-app message
- Privacy Protection: Notifications share only the general reason for concern without chat transcripts or details; reviewers aim to complete assessment within one hour
- Complementary Support: Trusted Contact supplements existing crisis hotlines and professional services; ChatGPT continues encouraging emergency services when appropriate
- Expert-Guided Development: Feature developed with input from 260+ licensed physicians, mental health clinicians, researchers, and organizations including the American Psychological Association
- User Control: Users can modify or remove their Trusted Contact anytime; Trusted Contacts can self-remove from the help center
- Broader Safeguards: ChatGPT includes additional protections: refusing self-harm instructions, de-escalating conversations, suggesting breaks during extended use, and directing users to localized crisis resources