OpenAI Trusted Contact: ChatGPT Safety Feature

Source: OpenAI Blog Date: 2026-05-07 URL: https://openai.com/index/introducing-trusted-contact-in-chatgpt/

Summary

OpenAI launched Trusted Contact, an optional safety feature allowing ChatGPT users to nominate a trusted adult who may be notified if automated systems and trained human reviewers detect self-harm discussions. Notifications undergo human review (target under one hour) before sending. Developed with 260+ physicians across 60 countries and an Expert Council on Well-Being and AI. Notifications are intentionally limited — no chat transcripts or conversation histories shared.

Implications

• Safety governance thread: first proactive safety notification system in a consumer AI product. If effective, this becomes the standard every vendor must match. The human-review requirement creates operational overhead but provides legal defensibility.
• Legal liability: the feature is a direct response to lawsuits over chatbot interactions and self-harm. The architecture (human review, limited data sharing, opt-in) is designed for court defensibility.
• Precedent for agent safety notifications: the pattern (automated detection → human review → limited notification) could extend beyond self-harm to other safety-relevant agent behaviors in enterprise contexts.