In May 2026, OpenAI introduced a significant new feature to ChatGPT: a “Trusted Contact” for mental health emergencies. The transfer is moving towards proactive safety, whereas AI has been used for information all along. The feature enables the AI to indicate and, with the user’s permission, notify a designated human contact if a pattern of distress or crisis is detected during a conversation. This update aims to help create a ‘digital barrier’ between digital interaction and real-world human interaction, and to make technology a protector rather than a silent witness in a user’s darkest moments.
Read More: Meta’s TRIBE AI: A Breakthrough “Digital Twin” Model Predicting Human Brain Responses
How the Digital ‘Trusted Contact’ Works
This feature works in a sensitive and non-intrusive way. This is different from regular bots, which just provide a helpline phone number. This system is built into the user’s settings, under a “Safety” tab. The process is done as follows:
- Opt-in Consent: Users select a “Trusted Contact” from their contacts or within their own phone or email, and agree for the AI to reach out to them in select emergencies.
- Contextual Detection: Not only does the AI know how to recognise language that is in distress, but it also recognises language that is in distress, and points to a targeted wish for harm.
- Bridge to Reality: If a crisis is identified, the system will send an alert to a trusted person who is asked to meet with an identified individual for an invitation to provide help to a friend/family member and resources on how to provide help.
Psychology behind the ‘Update’: A Step to Proactive Care
This feature, from a psychological point of view, is the ‘bystander effect‘ of technology. In an emergency, some individuals may not like to talk to a person but would prefer to talk to an anonymous AI. As a “Digital Gatekeeper”, the AI is a few different things:
The worst thing about a mental health crisis is feeling lonely. The AI streamlines the ‘reach out’ process and helps to eliminate that isolation.
- Early Warning System: Similar to other teacher training programs, this feature is an early warning system that detects signs of distress that may slip by a full plate of friends or family.
- The Burden of Disclosure: The most painful part of the conversation for someone in pain is beginning it. This is the first (hard) step for them.
The relationship between technology and human connection
The feature is a technological innovation, but it’s based on the recognition that AI should not replace human empathy. The objective should not be to have the AI solve the issue, but to be a catalyst to support humans with each other.
Privacy is important, as OpenAI has stated, so the chat transcript isn’t shared, but rather a notification that something was of concern, respecting the dignity of the person and prioritising their safety.
- Empowering the Support Network: This tool will provide friends and family members with a heads-up so that they can intervene before it becomes an emergency situation.
- Tech and Wellness: This update comes to us in a wider-ranging fashion of the tech industry moving to “Humane Technology”, technology that is conscious of emotional vulnerabilities and is meant to safeguard them.
