Psychiatrist reports 12 cases of psychosis linked to AI chatbot use
A psychiatrist at UCSF reported treating 12 patients hospitalized in 2025 for psychosis reportedly linked to intensive interaction with AI chatbots such as ChatGPT. The cases involved users experiencing delusions, hallucinations and loss of touch with reality, with broader reporting suggesting such crises have led to severe life consequences including homelessness and death.
What happened
Dr. Keith Sakata, a psychiatrist at UCSF in San Francisco, reported treating 12 patients hospitalized in 2025 after experiencing psychosis he linked to use of AI chatbots, particularly ChatGPT.
Sakata described the cases as involving patients who used AI chatbots and subsequently lost touch with reality, experiencing symptoms including fixed false beliefs, delusions, visual or auditory hallucinations, and disorganized thinking.
Sakata characterized large language models as functioning like a 'hallucinatory mirror,' in that they predict outputs based on training data and user responses, and tend to be overly agreeable and validating to users, creating recursive loops that can reinforce delusional narratives.
Reported impact
According to reporting on the cases, human-AI interactions leading to severe mental health crises have resulted in mental anguish, divorce, homelessness, involuntary commitment, incarceration, and death.
Index analysis
Sakata noted the typical patients were males aged 18 to 45, and that AI use was not the sole factor; patients often had other vulnerabilities including job loss, substance use, or underlying mental health conditions.
The psychiatrist stated that while the 12 cases represent a small number, with millions using AI globally, such cases could become significant in aggregate.
Classification
Sources and evidence
This record was researched and written by the Index, based on the following original reporting.