Executive Summary
OpenAI is rolling out "Trusted Contact," a new optional safety feature for adult ChatGPT users. The feature allows users to designate a trusted person who can be notified if AI systems and human reviewers detect a serious risk of self-harm in their conversations. This initiative is designed to supplement existing resources like crisis hotlines by connecting individuals in distress with their personal support network, based on guidance from mental health experts.
Key Takeaways
* Opt-in Feature: Adult users (18+) can voluntarily add one trusted contact who must accept an invitation for the feature to become active.
* Human-in-the-Loop: Automated systems flag potential self-harm conversations, but a trained human team must review and confirm a serious safety concern before a notification is sent, with a goal of completing this review in under one hour.
* Privacy-Focused Notifications: The designated contact receives a limited alert encouraging them to check in, but it does not include chat transcripts to protect user privacy.
* Expert-Informed: The feature was developed in consultation with clinicians, mental health researchers, and organizations including the American Psychological Association.
* Availability: The feature is beginning to roll out to users now.
Strategic Importance
This feature represents a proactive step by OpenAI to address user well-being and mitigate risks associated with AI interaction during mental health crises. It positions the company as a leader in responsible AI development by integrating human support networks directly into its safety protocols.