OpenAI has rolled out a new safety feature for ChatGPT focused on mental health. Called Trusted Contact, the tool is designed to notify a nominated person if the AI system, together with trained staff review, determines a user is engaging in conversations that indicate self-harm or other acute mental-health distress.
Availability and how it works
The feature is available to users aged 18 and older. To enable it, a user must provide the name of a trusted contact along with that person’s email address and phone number so the system can deliver a notification. The nominated contact receives an invitation from OpenAI explaining the feature and has one week to accept the request. If the contact declines, the user may designate a different person in their account.
Why OpenAI built it and what the data show
Trusted Contact expands ChatGPT’s existing safety alerts, such as notifications for parents when their child shows signs of anxiety. OpenAI says it developed the feature with input from health professionals, researchers, and mental-health organizations.
According to OpenAI data cited by Gizmodo, in 2025 signs of a “mental-health emergency” were observed in 0.07 percent of ChatGPT’s weekly-active users, while 0.15 percent presented a high risk of “self-harm or suicide.” An additional 0.15 percent showed indicators consistent with developing an “emotional dependency” on the AI. Given OpenAI’s estimate that roughly 10 percent of the world’s population uses ChatGPT weekly, these percentages could translate to nearly three million people.
Further guidance and context
OpenAI emphasizes that ChatGPT continues to encourage users to contact crisis hotlines or seek medical help when needed. In Hungary, crisis lines reachable free of charge are 116‑123 and 06 80 820 111.
Similar contact-notification features exist on other platforms, including Instagram; OpenAI says the Trusted Contact rollout aims to provide earlier warning and help prevent tragic outcomes.
