OpenAI expands AI safety measures with ‘Trusted Contact’ feature

OpenAI has unveiled a significant new feature in ChatGPT aimed at bolstering mental health support. The 'Trusted Contact' feature allows users to connect with a friend or family member for assistance during critical moments. When ChatGPT identifies discussions related to self-harm, it will prompt users to reach out to their designated trusted contact, ensuring an additional layer of human support when it is most needed. So, how does this innovative feature work? Users are empowered to select one trusted adult—be it a friend or family member—who can be notified if ChatGPT detects concerning conversations. OpenAI emphasizes the importance of facilitating real-world connections during challenging times. The process involves four straightforward steps, from selecting a contact to sending notifications. To set up the 'Trusted Contact' feature, users start by choosing one trusted individual as their emergency contact within ChatGPT. They can send invitations via email, SMS, or other messaging platforms. The chosen contact must accept this invitation within one week; otherwise, users will need to select someone else. If ChatGPT's monitoring systems flag conversation topics related to suicide or self-harm, trained human reviewers will evaluate the discussions for any potential risks. Should a serious threat be identified, the trusted contact will receive a notification encouraging them to check in on the user. Importantly, OpenAI has made it clear that no chat details or conversation records will be shared with the trusted contacts. Users are advised to choose someone they feel comfortable with and trust to respond thoughtfully. When sending the invitation, ChatGPT provides clear information about its role and the nature of the safety notifications to ensure transparency and understanding.