recently, openai launched a new feature for chatgpt called “trusted contacts,” designed to proactively intervene and provide interpersonal support when users’ conversations contain content related to self-harm. this feature is primarily aimed at adult users: once the system detects potential risks of self-harm or suicide, it encourages users to reach out to someone they trust while automatically sending alerts to pre-set contacts, urging them to offer extra care and communication.
prior to this, openai had faced multiple lawsuits in which plaintiffs alleged that chatgpt’s responses encouraged their loved ones to commit suicide or even assisted in crafting suicide plans. to address this sensitive issue, openai now employs a combined approach of automated screening and human review. when a conversation triggers a risk alert, the relevant information is forwarded to the human safety team. according to official statements, every such notification undergoes manual review, with efforts made to complete the assessment within one hour.
if the internal assessment determines that a user faces a severe risk of self-harm or suicide, chatgpt will send an email, text message, or push notification to the user’s designated “trusted contact.” these notifications are brief and primarily encourage the contact to initiate communication with the individual; to protect user privacy, they do not include any specific details of the conversation. the goal is to transform cold ai interactions into genuine human connections, offering an extra measure of warmth and support at critical moments.
