OpenAI adds Trusted Contact alerts to ChatGPT for moments of crisis

OpenAI has launched Trusted Contact, a new optional safety feature for ChatGPT. It lets adult users name trusted contacts — relatives, friends or guardians — who will be notified if the system detects signs of a psychological crisis.

How it works

The feature is simple: the user specifies one or more trusted contacts and their contact information. If ChatGPT detects that a person is discussing self-harm or suicide, the system sends a notification to those contacts.

OpenAI built the feature on expert analysis: research shows that when a person is in acute crisis, support from someone they know and trust can be critically important. It is not a replacement for professional help, but a complement to it. Trusted Contact is added as another layer of safety on top of existing mechanisms. When a user mentions suicide or self-harm, ChatGPT suggests local phone hotlines (for example, 988 in the US). Now the ability to alert loved ones is added to that.

Opt-in by design

The feature is not mandatory. Users decide for themselves whether to turn it on, whom to name as trusted contacts and when to turn it off. This approach is critical for trust: mental health is an intimate area, and not everyone will accept automated monitoring, even for their own safety.

• The user has full control over the list of trusted contacts
• Notifications are sent only when a conversation about a crisis is detected
• The feature can be turned off at any time

Contacts will receive not just an alert but information about what happened, along with a list of resources for offering support. The idea is to give people a tool, not cause panic.

Context: the responsibility of AI companies

OpenAI has previously worked on how AI systems should respond to mentions of suicide and self-harm. The balance is delicate: on one hand, the system needs to help a person in distress; on the other, it must not intrude on privacy or create the false impression that AI can replace real help.

Many companies in this space are experimenting with safety features. Some use chatbots for screening, others integrate hotlines. Trusted Contact sits somewhere in the middle: not total control, but an expanded support network.

"Trusted Contact is built on a simple, expert-backed idea: when someone may be in crisis, connecting with someone they know and trust can make a real difference,"

OpenAI writes.

What this means

This reflects the growing responsibility of AI companies for user well-being. The feature does not claim to be a miracle cure, but it is one more tool in the arsenal — a bridge between technology and human care, between a chatbot and people who are genuinely ready to help.