OpenAI yesterday unveiled a new safety feature for ChatGPT that’s unlike anything we’ve seen in the industry: Trusted Contact. Users can designate a person they trust who may be notified if automated systems and trained reviewers detect the user might be discussing self-harm.
How it works
You can add an adult as your Trusted Contact in ChatGPT’s settings. That person receives an invitation they need to accept within one week.
If ChatGPT’s automated monitoring systems detect potential self-harm, here’s what happens: ChatGPT tells the user it may notify their Trusted Contact and suggests conversation starters for reaching out. A specially trained team reviews the situation. If the reviewers determine there may be a serious safety concern, the Trusted Contact gets a brief notification via email, text message, or in-app notification.
The double-edged sword
The feature is entirely voluntary — nobody gets monitored automatically. But it still raises important questions.
On one hand: AI chatbots have become a conversation partner many people turn to in difficult moments. Having a safety net makes sense.
On the other hand: an AI system that analyzes conversations for self-harm and notifies third parties is treading on sensitive ground. Where’s the line between care and surveillance?
What OpenAI promises
According to OpenAI, the Trusted Contact only receives a brief notification — no conversation content, no details. The feature is meant to be an additional safety net, not a surveillance tool.
The bigger picture
After the FSU shooting case and the ongoing criminal investigation in Florida, OpenAI is under massive pressure to do more for user safety. Trusted Contact is one response — not perfect, but a first step in a direction that other AI companies will need to follow as well.
Sources: OpenAI: Introducing Trusted Contact in ChatGPT, TechCrunch: OpenAI introduces new ‘Trusted Contact’ safeguard