Strengthened Protection for ChatGPT Users: Trusted Contact Alerts for Detected Self-Harm Risks

Notification Sent to "Trusted Contact" via Text Message, etc.
Original Conversation Not Shared

A new feature will be introduced to ChatGPT that allows users to notify a trusted individual if serious self-harm risk signals are detected during conversations with the AI.

On May 10, OpenAI announced that it will add a "Trusted Contact" feature to ChatGPT. With this feature, adult users can register one trusted person—such as a family member, friend, or guardian—in advance. If ChatGPT detects serious concerns about self-harm, it can help connect the user to real-world support through the pre-registered contact. Users can register one trusted person in ChatGPT’s settings, and in Korea, only adults are eligible to register. This feature is based on the "Safety Alert" system from ChatGPT’s parental control features.

If an expression indicating a serious risk of self-harm is detected during a conversation, the system informs the user that ChatGPT may notify the trusted contact and provides sample dialogue for the user to reach out directly. Afterwards, a specially trained reviewer will reassess the situation, and if a real safety concern is determined, a notification will be sent to the trusted contact via email, text message, or app notification. However, the notification will not include the original conversation or its content.

Users can modify or delete their "Trusted Contact" at any time, and the registered contact information can be disconnected whenever desired. The trusted contact can also request removal at any time.

OpenAI stated that this feature was designed based on consultations with its global network of experts and a well-being and AI advisory committee. The company also collaborated with external professional organizations including the American Psychological Association (APA).

Giseok Ko, OpenAI’s policy chief, explained, "People use ChatGPT to solve problems and reflect on personal concerns, but sometimes users might be going through difficult moments. This feature was introduced based on the belief that AI systems should be designed to help connect users to real-world support in such situations."

Meanwhile, on April 24, OpenAI was selected as a "Cheonmyeong Guardian Center" for the "2026 Cheonmyeong Guardian Project," which aims to promote a culture of respect for life and lead suicide prevention efforts. This project is led by the Government-wide Life Keeper Promotion Headquarters under the Office for Government Policy Coordination, with the goal of reducing the number of suicide deaths by 1,000 this year.

This content was produced with the assistance of AI translation services.