ChatGPT adds trusted-contact alert to notify a nominated contact in crisis situations

ChatGPT now has a feature that can contact a real person if a conversation looks like a self-harm emergency. That is the actual news here — not a vague “better safety” promise, but a specific product setting called Trusted Contact that OpenAI started rolling out on May 7. Adult users with personal ChatGPT accounts can choose one person they trust, and if OpenAI’s systems think the user may be at serious risk of suicide, that person may get an alert. The point is simple: get one human into the loop before things get worse. (openai.com) ### What is Trusted Contact? It’s an optional safety setting inside ChatGPT. You pick a trusted person, ChatGPT sends them an invitation, and nothing happens unless that person accepts. OpenAI says the feature is for adults with personal accounts in supported regions, not for Business, Enterprise, or Edu workspaces. (help.openai.com)on of sadness or distress triggers this. OpenAI says the threshold is much narrower — cases where automated systems and trained reviewers detect talk about suicide that signals a serious safety concern. If that happens, ChatGPT may notify the trusted contact and encourage them to check in. OpenAI also says ChatGPT will still push users toward crisis lines or emergency services when needed. (openai.com) ### Why add a human reviewer? Because this is the part where false alarms really matter. A model can misread tone, sarcasm, fiction, or venting. OpenAI built a two-step gate here — automated detection first, then trained human review before an alert may go out. Basically, the company is trying to avoid the worst version of this feature, where a bot panics and emails your friend because you had a dark joke in a late-night chat. (help.openai.com) ### Who can get the alert? Just the one person the user chose in advance. This is not a system that contacts police, hospitals, or random emergency contacts by default. The feature is framed as a way to strengthen social connection, which mental-health experts often treat as a protective factor in suicide prevention. OpenAI is pretty explicit that this does not replace professional care. (openai.com) ### Why is OpenAI doing this now? Because ChatGPT has drifted into a role a lot of people never formally assigned to it — part assistant, part confidant, part always-awake listener. OpenAI has been tightening its mental-health safeguards for months, including stronger responses around suicide, psychosis, mania, and emotional dependence, plus parental safety notifications for teens. Trusted Contact is the next step in that same direction. (openai.com) ### What’s the privacy catch? The catch is that a private chat can now lead to a real-world notification if the risk looks serious enough. Even with consent and review, that changes the feel of the product. Users need to understand that enabling the feature means some conversations could cross the boundary from “chat with AI” to “AI prompts a human intervention.” OpenAI’s setup tries to make that boundary explicit by making the feature opt-in and invitation-based. (help.openai.com) ### Is this replacing crisis support? No — and OpenAI keeps saying that directly. ChatGPT can surface crisis resources, but it is not a therapist, a hotline, or emergency response. In the U.S., the company points people to 988 for immediate suicide and crisis support. Trusted Contact is more like a backup rope than a rescue team. (openai.com)t does more than talk — in rare cases, it can now try to pull another person into the room. That could help. But it also means the line between private AI companion and safety system just got a lot more real.