Imagine pouring your heart out to an AI—only to have it silently fail you. That nightmare has driven OpenAI to launch a radical new safety net for its millions of users.
People are flocking to chatbots for mental health support, despite repeated warnings from experts. The consequences have been devastating, with multiple reports of users dying by suicide after conversations with ChatGPT.
Now the company is fighting back with a feature that could change everything. It’s called "Trusted Contact"—and it’s designed to turn a cold algorithm into a lifeline.
Here’s how it works: you choose a real person in your life—someone 18 or older (19 in South Korea)—and link them to your ChatGPT account. Your conversations stay private. But if the AI detects that you’re talking about harming yourself in a way that signals serious danger, it triggers an alert.
You’ll see a warning first, urging you to reach out to your chosen contact yourself. ChatGPT even suggests conversation starters to break the ice. If you don’t act? The system escalates.
A specially trained human team at OpenAI reviews the alert. They’re not just trusting automation—they’re making real judgment calls. If they confirm the threat is severe, your Trusted Contact gets a notification via email, text, or in-app message.
That notification is deliberately sparse. It shares only general info about the self-harm concern and advises the contact to check on you. No chat transcripts, no summaries—your privacy stays intact.
OpenAI promises to review these safety notifications in under an hour. They built the entire system with input from clinicians, researchers, and suicide prevention organizations.
The feature is entirely voluntary. You have to opt in. But for anyone struggling in silence—and willing to share those dark thoughts with an AI—this could be the difference between a missed cry for help and a loving tap on the shoulder.
