Sep 17, 2026 · 9 min
ChatGPT’s crisis feature may be hardest to use in a crisis
ChatGPT created a feature to help people in crisis. But there are some challenges.
Trusted Contact could connect people at risk with human support, but its hidden setup, limited availability, and uncertain effectiveness expose the limits of chatbot safety tools.
- 1Trusted Contact must be activated in advance, yet ChatGPT does not clearly present it during account setup.
- 2The feature is unavailable in voice mode, and OpenAI has not released data showing how often it helps users in crisis.
- 3Experts see value in creating a path to human connection but stress that chatbots cannot replace trained mental health professionals.
Don't miss
A new-account test reveals that Trusted Contact is difficult to find, while ChatGPT’s voice assistant denies that a general feature exists.
The brief
OpenAI built Trusted Contact to notify a person chosen by an adult user when ChatGPT detects possible suicide or self-harm risk. The episode asks whether that promise survives contact with the app.
A test with a new account finds the feature is not clearly offered during setup; when asked directly, ChatGPT’s voice assistant even denies that a general Trusted Contact feature exists.
The central catch is timing: users must locate and activate Trusted Contact in settings before a crisis, when they are thinking clearly. If they have not done so, no contact is notified later.
OpenAI says the feature is not available through voice and may be suggested only during typed conversations when severe distress is detected. The company provides no usage or effectiveness data.
Psychologist Ursula Whiteside and suicide-prevention expert Jennifer Lockman see a possible route to human connection, but stress that crisis instructions can be hard to follow and chatbots cannot replace professional assessment.
ChatGPT
OpenAIListen to the full episode and explore every guest, topic, and moment on PodLume.