OpenAI adds a Trusted Contact alert feature to ChatGPT's safeguards

OpenAI is expanding its efforts to protect ChatGPT users in situations where conversations may touch on the risk of self-harm. The company has introduced a new feature, Trusted Contact, which can automatically alert a trusted person to a potential threat.

How Trusted Contact works

The feature lets users name a contact in advance: a person who should be notified if signs of self-harm risk are detected in a ChatGPT conversation. If the system determines that the user is discussing possible self-harm or suicidal thoughts, it sends the trusted contact a notification about the potential danger.

The mechanism uses OpenAI models to analyze text in real time. The notification describes the risk but does not include the full context of the conversation, to protect the user's privacy. The trusted contact receives a targeted alert along with links to support resources.

The feature is entirely optional: users must activate it themselves and choose whom to trust with the notifications. This means the system will not send alerts without explicit consent.

Why this matters for platforms

Users' mental health is a growing concern for companies that build AI systems. Research shows that people in crisis often turn to ChatGPT with questions they are not ready to discuss with friends, parents or doctors. Anonymity and the absence of judgment make an AI assistant appealing for such conversations. OpenAI is trying to balance the accessibility of the service against its responsibility to vulnerable users.

Trusted Contact is not the first mechanism of its kind. The company already offers integration with mental health helplines, warning messages in critical situations and information about support resources.

The approach fits global trends. Other large platforms (Meta*, TikTok, Discord) are also developing tools to detect risk and alert loved ones. It is becoming standard practice in the industry.

Where this fits in OpenAI's strategy

OpenAI is investing in several areas of user protection:

• Automatic detection of critical situations in dialogues through ML models
• Built-in links to hotlines, counselors and support communities
• Educational materials on mental health and available resources
• Notifications and recommendations for parents of underage users
• Partnerships with aid organizations (NAMI, SAMHSA, Crisis Text Line)

Trusted Contact is designed specifically for users who recognize the risk and want a safety net, so that a loved one learns about the problem early and can offer support.

What this means

Trusted Contact shows AI companies taking growing responsibility for user safety. The mechanism does not replace professional help, but it can save time at a critical moment, when a trusted person learns of the problem and can step in. For the tens of millions of ChatGPT users, it adds one layer of protection to the platform's ecosystem.

*Meta has been designated an extremist organization and is banned in Russia.