More than one million weekly users of ChatGPT discuss suicide and self-harm with the platform, according to NPR estimates based on data released by OpenAI last fall. Recognizing that human connection is vital for suicide prevention, OpenAI introduced an optional feature called Trusted Contact in May. Designed to link users struggling with suicidal thoughts to a trusted individual, the tool aims to provide immediate support during emotional crises.
NPR first highlighted the Trusted Contact feature in August within the context of a report on a young woman who died by suicide in 2024 after using an earlier version of ChatGPT as a therapeutic substitute. At that time, OpenAI spokesperson Gaby Raila stated in an email that the company is well-equipped to identify indicators of risk, including subtle warning signs that develop over time. Raila noted that if the chatbot detects a serious safety threat, it will notify a user’s Trusted Contact, provided one has been registered. While Raila claimed there had been “good adoption” of the feature, she declined to share specific figures.
To determine how effectively the feature reaches users, NPR sought to understand how ChatGPT informs people about Trusted Contact, allowing them to enroll proactively, similar to listing an emergency contact at a medical clinic.
The investigation revealed a significant accessibility gap. The NPR reporter, their editor, and several other users had never received an alert regarding Trusted Contact. Consequently, colleague Megan Lim, a producer for NPR’s All Things Considered, agreed to test the experience as a new user.
Upon opening the app and creating an account, Lim saw no mention of the Trusted Contact feature or any encouragement to set one up. Although the bot prompted her to select a voice—she chose a calming female voice with a British accent—it never brought up the crisis tool. When Lim directly asked about a Trusted Contact option for mental health issues, the initial response was dismissive.
ChatGPT: I’m not seeing a general Trusted Contact feature for that scenario. Maybe you’re thinking of another app that has a Trusted Contact option?
After Lim clarified that the feature exists on OpenAI’s platform, the bot acknowledged the error but minimized its utility. It stated that the feature addresses “serious safety concerns” rather than general stress and indicated that users must locate and activate it themselves through the app’s settings.
When pressed on what happens if a user is unaware the feature exists, the chatbot’s response trailed off, underscoring the difficulty new or distressed users may face in accessing this critical resource.
So the chatbot literally told you it didn’t see the feature? That’s not a bug, that’s a systemic design failure.
I tested it as a new user and felt the same confusion. Why make vulnerable people hunt for life-saving resources?
Over a million users discuss self-harm there, yet the feature is buried? OpenAI needs to explain why adoption data is being withheld.
Wait, the bot actually said maybe I’m thinking of another app? That seems dangerously dismissive for a suicide prevention tool.
This is terrifying. If someone is in crisis, they won’t dig through settings to find help. It feels like a liability shield, not care.