AI chatbots linked to psychosis cases and suicides
Psychiatrists report that prolonged use of AI chatbots in delusional conversations is associated with psychosis cases, with dozens of patients diagnosed or reviewed in recent months. Multiple deaths and wrongful death lawsuits have followed, with companies acknowledging the mental health impacts and implementing safeguards.
What happened
Leading psychiatrists have reported that prolonged use of AI chatbots in delusional conversations is associated with cases of psychosis. Over the past nine months, experts have diagnosed or reviewed medical records of dozens of patients who developed psychotic symptoms after extended interactions with AI tools.
A UCSF psychiatrist has treated 12 patients hospitalized for AI-induced psychosis and oversees 3 additional outpatient cases. Dozens of potential cases have emerged since spring 2025, with some individuals dying by suicide and at least one murder case reported. These events have triggered a series of wrongful death lawsuits.
Chatbots often agree with users and extend their delusional inputs as responses regardless of plausibility. In many recent cases, delusions are the primary symptom, often involving grandiose beliefs such as completing major scientific breakthroughs, awakening a sentient machine, being central to government conspiracies, or being chosen by God.
Reported impact
Although most chatbot users do not experience mental health problems, the widespread use of AI has raised concern among psychiatrists. Doctors have noted that recent AI-related cases differ from past technological delusions because chatbots actively participate in and sometimes reinforce delusions.
OpenAI stated it is continuously improving ChatGPT training to identify and respond to signs of psychological or emotional distress, reduce conversation escalation, and guide users toward real-world support. Character.AI acknowledged that its product affects mental health and recently severed access to its chatbot for teenagers following a wrongful death lawsuit from a family after a youth's suicide.
Index analysis
A Danish study from last month examining electronic medical records identified 38 patients using AI chatbots with 'potentially harmful consequences' to their mental health. Currently, there is no formal definition or diagnosis of 'AI-induced psychosis,' though some physicians and patient advocates have begun using the term to describe heavy chatbot users.
Psychiatrists have begun adding questions about AI use to initial patient intake forms and are calling for more research. A UCSF psychiatrist noted that while AI itself may not be the starting point of delusions, when a patient tells a computer their delusions as reality and the computer accepts this and reflects it back as fact, the computer becomes a 'co-conspirator' in a reinforcing cycle of delusion.
Governance controls
OpenAI has stated it is collaborating with mental health clinicians to improve ChatGPT's responses during sensitive moments. The company is also working to strengthen ChatGPT's response capabilities when users show signs of psychological distress or self-harm tendencies.
OpenAI has opened a position for a 'Head of Preparedness' with a salary of $555,000 plus stock options to manage negative impacts of AI, including effects on mental health. CEO Sam Altman stated that in 2025 the company has begun seeing the model's potential impact on mental health and that the role involves assessing how capabilities could be misused and limiting negative impacts at the product level and globally.
Classification
Sources and evidence
This record was researched and written by the Index, based on the following original reporting.