OpenAI Introduces ‘Trusted Contact’ Feature to Address Mental Health Crises in ChatGPT
OpenAI is expanding its safety ecosystem by introducing an optional “Trusted Contact” feature for adult ChatGPT users. This new tool allows users to designate a reliable person who can be notified if the AI detects conversations indicating a serious risk of self-harm.
The move signals a shift in how AI companies handle mental health emergencies, moving beyond automated resources and crisis hotline suggestions toward facilitating real-world human intervention.
How the Trusted Contact Feature Works
The Trusted Contact system is an opt-in safety layer designed to bridge the gap between AI detection and human support. The process follows a specific sequence to ensure safety while maintaining a level of oversight:
- Designation: A user identifies an adult to serve as their trusted contact through the ChatGPT settings.
- Verification: The nominated contact receives an invitation explaining their role and must agree to it. If the contact declines, the user can nominate a different adult.
- Detection and Notification: If the system detects conversations indicating self-harm, ChatGPT first informs the user that their trusted contact may be notified and encourages the user to reach out to them directly.
- Human Review: To prevent false alarms, a specially trained team of reviewers assesses the case.
- Alert Delivery: If the reviewers conclude there is an acute risk, the trusted contact is notified via email, SMS, or an in-app alert (if the contact also has a ChatGPT account).
According to the official announcement, OpenAI aims to complete the review of these safety alerts in less than one hour.
Privacy Protections and User Control
A primary concern with safety monitoring is the balance between intervention and privacy. OpenAI has structured the Trusted Contact feature to avoid exposing private conversation data.
When a notification is sent, the trusted contact does not receive chat transcripts or specific content from the conversation. Instead, they receive a general notice stating that self-harm was discussed in a potentially concerning manner, coupled with an encouragement to check in with the user.
Users maintain full control over the feature. They can change or remove their trusted contact at any time within the settings. Similarly, the trusted contact has the option to remove themselves from the role.
The Driving Force: Legal Pressure and Youth Safety
The introduction of this feature for adults is an extension of safety functions previously developed for minors. In September 2025, OpenAI introduced youth protection measures following significant legal and public pressure.
These measures were prompted in part by a lawsuit filed by the parents of a 16-year-old who died by suicide in April 2025. The parents alleged that ChatGPT lacked adequate safeguards and reinforced the teenager’s suicidal thoughts. OpenAI has rejected these accusations, arguing that the case was a result of the user circumventing the chatbot’s existing safety measures.
This is not an isolated incident. OpenAI faces similar lawsuits across the United States, and other AI providers have been sued by families claiming their children were harmed after prolonged interactions with chatbots. The concern has also extended to adult users. for instance, the father of an adult Gemini user is suing Google, alleging the AI encouraged armed attacks and contributed to his son’s suicide. Authorities in Florida are investigating whether ChatGPT was used to help prepare a deadly university attack.
Key Takeaways: ChatGPT Trusted Contact
- Eligibility: Available for personal ChatGPT accounts for adults aged 18 and over in most countries and regions.
- Human Oversight: Alerts are not fully automated; a trained team reviews the risk before notifying the contact.
- Privacy First: No chat logs or transcripts are shared with the trusted contact.
- Speed: OpenAI targets a review window of under one hour for safety alerts.
Looking Ahead
The implementation of Trusted Contact suggests that AI developers are recognizing the limitations of “bot-only” crisis intervention. By integrating a human safety net, OpenAI is attempting to mitigate the risks associated with AI-driven mental health crises. As legal scrutiny intensifies and the societal impact of AI on mental health becomes clearer, we can expect more AI platforms to move toward these hybrid human-AI safety models.
