OpenAI is rolling out a new safety tool called Trusted Contact to help users during mental health emergencies. This optional feature lets the platform reach out to a friend or family member if the system sees a risk of self-harm in a chat. It works as an extra layer of support for adults. The company wants to make sure that someone you trust gets a message when the software sees a high risk of danger. By using both AI and human reviewers, the company hopes to help users get real life support from people they know.
How the Safety Alert System Works
The system follows a few steps. First, automated tools scan your chats for any language about self-harm. If the AI sees a problem, it does not just message your contact immediately. It first shows you a message with links to professional help. Honestly, I think this is a good way to give the user a chance to reach out themselves first. It also suggests ways to start a talk with a friend.
Actually, let me put that more simply. A team of trained human reviewers looks at the flagged chat after the AI picks it up. These specialists check the context to see if the danger is real. If the situation is considered serious, the platform sends a short notification to the chosen contact. This message can be sent as an email, text message, or as a notification in the ChatGPT app, if the contact has an account. Wait, before we get into the data part, you should know this whole process usually happens within one hour.
Privacy and User Control Over Data
Privacy is a big deal here. OpenAI says the note sent to your friend does not show your actual chat. Your trusted contact will never see what you wrote to the AI. Instead, the alert just says the system found a safety concern. It also gives the contact some professional resources to help them support you.
You have full control. You must go into your settings to add a person. When you invite someone, they have seven days to accept the request. If they say no or do not reply, the feature stays off. You can change your person or turn the whole thing off whenever you want. Basically, the power stays in your hands.
Expert Guidance and Who Can Use It
OpenAI did not make this alone. They worked with doctors and mental health groups like the American Psychological Association. They also use the Global Physicians Network. This group has over 260 doctors from 60 countries who give feedback on how the AI talks about health.
Look, this is for adults only. In most places, you and your contact must be at least 18 years old. But in South Korea, the age limit is 19 to follow their local laws. This new tool joins other safety steps the AI already uses. For example, if you ask about self-harm, the bot already gives you helpline numbers. The Trusted Contact feature goes further by involving a human support system.
I feel this is a very thoughtful addition to the app. Most of us use AI to vent or talk about our day. Having a safety net like this makes the tech feel a bit more human. We will have to see how well the human reviewers handle the local context in India. It is a big step for a tech company to take.
