OpenAI has introduced an important new safety feature for ChatGPT users called “Trusted Contacts.” The artificial intelligence system will now be able not only to provide advice, but also to alert a user’s close contacts if conversations show signs of a psychological crisis or dangerous behavior.

How does it work?

The system is activated only in exceptional situations when the AI detects signals indicating a risk of self-harm. The process works as follows:

• Warning: ChatGPT first encourages the user to seek help on their own.

• Review: The alert is sent to a specialized OpenAI safety team to verify whether the threat appears genuine.

• Notification: If the risk is confirmed, a brief message is sent to the trusted contact via SMS, email, or an app notification.

Privacy and data protection

An important aspect of the feature is maintaining conversation privacy. Trusted contacts will not gain access to chat history, screenshots, or direct quotes. They will receive only a general notification about a possible safety concern involving the user. This approach helps balance privacy with the need for emergency support.

The feature is available on a voluntary basis for adult users. To activate it, users must provide the contact details of a trusted person and receive their confirmation within one week. Users retain full control and can change the trusted contact or disable the feature at any time. Developed in cooperation with experts, the new tool is designed to support people during critical moments.

ORIENT