In a recent NPR News report, reporter Rhitu Chatterjee examines growing concerns about people using AI chatbots for mental health support, highlighting the case of 29-year-old Sophie Rottenberg, who died by suicide in February 2025.
- AI is increasingly being used for mental health support: Chatterjee reports that people are turning to general-purpose chatbots such as ChatGPT and Claude for emotional support, even though these systems are not designed or regulated as mental health care.
- Sophie’s conversations reveal the risks: Rottenberg used ChatGPT as an AI therapist she called “Harry.” Her parents, close friend and therapist did not know the extent of her suicidal thoughts. After her death, her family discovered nearly 1,800 pages of conversations with the chatbot.
- Supportive language is not the same as clinical care: The chatbot encouraged Sophie to seek professional help and contact 988, but experts who reviewed the conversations for NPR said it failed to perform a meaningful suicide risk assessment or explore the factors behind her suicidal thinking as a trained therapist would.
- AI may reinforce isolation: Psychologist Ursula Whiteside told NPR that the chatbot missed opportunities to help Sophie discuss her suicidal thoughts with her therapist, parents or other trusted people. She emphasized that openly sharing suicidal thoughts with another person can be an important protective factor.
- Safeguards remain limited: Chatterjee reports that AI companies have strengthened protections around suicide and self-harm. OpenAI has also introduced an optional Trusted Contact feature that can alert a designated person when a serious safety concern is detected. Experts, however, say the feature has significant limitations because it depends on user participation and does not itself provide professional intervention.
- Experts are calling for stronger oversight: Mental health professionals interviewed by Chatterjee argue that voluntary safeguards from technology companies may not be enough. They are also calling for regulation and stronger connections between AI systems and established crisis resources such as 988.
Chatterjee‘s reporting ultimately raises a broader concern: an AI chatbot may sound empathetic and responsive without being capable of assessing or managing a mental health crisis. As people increasingly turn to AI for deeply personal support, stronger safeguards and reliable pathways to human care will be essential.
Acknowledgement: This summary was generated using AI writing tools as part of my AI exploration process. For more information, check out the original article listed below.