Asger Risom
OpenAI has launched a new ChatGPT safety feature that can alert a trusted friend or family member if human reviewers determine a user may be at serious risk of self-harm or personal danger.
OpenAI has introduced a new ChatGPT safety feature that allows users to designate a trusted contact who can be alerted if the AI detects signs of self-harm or serious danger.
The system adds a new layer of human intervention to ChatGPT’s mental health safeguards, with OpenAI saying trained reviewers will assess high-risk conversations before any notification is sent.
New safety feature
The new “Trusted Contact” tool allows adult users to nominate a friend, family member or caregiver through ChatGPT’s settings.
According to elEconomista, the feature is designed for situations where users turn to ChatGPT while struggling emotionally or discussing personal crises.
If OpenAI’s automated systems detect conversations involving potential self-harm or threats to personal safety, the interaction may be flagged for review by trained human specialists.
Should reviewers determine there is a serious risk, ChatGPT can send an alert to the user’s chosen contact through email, text message or an in-app notification.
OpenAI said the goal is to create “an additional level of support” beyond the crisis helplines and emergency guidance already integrated into the chatbot.
Human review involved
The company stressed that notifications are not triggered automatically by AI alone.
Instead, conversations identified by automated systems are reviewed by human teams before any warning is sent, in an effort to reduce false alarms and assess the seriousness of the situation.
The designated contact must be over 18 and must actively accept the role within one week after receiving an invitation from the user.
If the invitation is declined, the user can choose another trusted person instead.
OpenAI said the feature builds on existing parental safety tools already used for younger users on the platform.
Growing AI concerns
The rollout comes as AI companies face increasing scrutiny over how chatbots are used during emotional distress, loneliness and mental health crises.
ChatGPT has become widely used not only for work and research, but also for emotional support, personal advice and private conversations.
OpenAI said the assistant will continue encouraging users to contact emergency services, crisis hotlines and professional support services where necessary.
The company also emphasized that the trusted contact feature is not intended to replace medical or psychological care.
The announcement reflects a broader debate inside the tech industry over how much responsibility AI platforms should carry when users discuss self-harm, depression or dangerous behaviour.
Sources: elEconomista
