ChatGPT’s Optional Trusted Contact Feature for Suicide Prevention

More than a million weekly users of ChatGPT engage in conversations about suicide and self-harm, according to NPR estimates based on data released by OpenAI. Suicide prevention experts emphasize the importance of human connection in preventing suicide. In May, OpenAI launched an optional feature called Trusted Contact to help connect individuals experiencing thoughts of suicide and self-harm to someone they trust during emotional crises.

NPR initially reported on Trusted Contact in August, focusing on a case involving a young woman’s suicide after using an older version of ChatGPT. OpenAI spokesperson Gaby Raila confirmed that ChatGPT is equipped to identify signs indicating a user may be at risk, based on both direct and indirect cues that appear over time. If the chatbot detects a serious safety risk, it will notify the user’s Trusted Contact, if one is listed. Raila noted the feature’s good adoption but did not provide specific user numbers.

NPR investigated how ChatGPT informs users about setting up the Trusted Contact feature. They encouraged users to set it up in advance, similar to listing an emergency contact at a doctor’s office. However, many users, including NPR personnel, had not been alerted about Trusted Contact. NPR’s Megan Lim tested the feature by opening the app for the first time, but the app did not mention Trusted Contact until Megan directly inquired. ChatGPT acknowledged the feature, explaining it is available for serious safety concerns, not general stress, and users need to opt-in through the app’s settings.

Raila mentioned that ChatGPT would advise users expressing distress to reach out to loved ones or contact the National Suicide and Crisis Lifeline (988). While Trusted Contact is currently unavailable on the voice version of ChatGPT, OpenAI is working on integrating this feature. If the chatbot detects suicidal thoughts during typing, it may suggest setting up Trusted Contact. Raila explained that without prior setup, ChatGPT can’t notify anyone, although it provides supportive responses.

“What they thought would be helpful for them in the future would be to have this trusted contact option,” said Ursula Whiteside, CEO of Now Matters Now.

Experts like Holly Wilcox from the Center for Suicide Prevention at Johns Hopkins advocate proactively offering the feature to all users before emergencies arise. Psychologist Ursula Whiteside evaluated the feature, finding it straightforward but challenging for users under extreme distress. Whiteside’s non-profit works with individuals struggling with suicidal thoughts and attempts who also use chatbots for support. Focus groups conducted earlier this year indicated a desire for a trusted contact feature, allowing users to establish connections when thinking clearly.

Psychologist Jennifer Lockman, from the University of Alabama’s suicide prevention clinic, expressed skepticism about the feature’s effectiveness in helping users overcome personal barriers to reaching out to contacts. Lockman highlighted self-perceptions among suicidal individuals, who may feel burdensome to others. She emphasized that unlike trained mental health providers, chatbots can’t replicate contact identification processes.

Despite reservations, Lockman acknowledged the feature’s potential to promote human connection, a critical protective factor in suicide prevention. Raila reiterated OpenAI’s collaboration with psychologists and researchers to refine when and how ChatGPT introduces Trusted Contact. OpenAI employs additional safeguards, encouraging users to take breaks and informing them about crisis hotlines like 988. Lockman remains hopeful about the feature’s potential impact but insists on research to determine its effectiveness in saving lives.

If you or someone you know is in crisis or contemplating self-harm, call or text the Suicide & Crisis Lifeline at 988.