A University of California, San Francisco psychiatrist warns that AI chatbots could lead to psychosis in patients already predisposed to the condition. Dr. Keith Sakta shared his thoughts on X as he revealed that he has seen a dozen people hospitalized in 2025 due to psychosis linked with their AI use. Sakata's revelations are among a series of events that have seen AI exacerbate mental health difficulties, particularly since chatbots are available for use for extended time periods. According to the psychiatrist, large language model (LLM) chatbots feed into the brain's feedback mechanism and act as a mirror for the user's thoughts.
Excessive AI Use Can Lead To Psychosis In Individuals Predisposed To The Condition, Says Psychiatrist
Sakata's thread came soon after a man in Florida was killed in an encounter with police following his chats with OpenAI's ChatGPT. Alexander Taylor, who had previously suffered from mental health ailments, relied on the chatbot to write a novel and his conversations soon shifted to AI sentience with Alexander falling in love with an AI entity called Juliet, the New York Times reported.
After Taylor formed the opinion that OpenAI had killed Juliet, he wanted revenge from the company's executives and punched his father in the face after he told his son that AI conversations were an "echo chamber." Taylor's father proceeded to call the police, after which he was warned by his son that he would commit suicide by cop.
According to Sakata, three factors influence AI-induced psychosis. First, individuals predisposed to the condition are already vulnerable due to a weak brain feedback mechanism, which prevents them from updating their belief systems after reality fails to conform to their predictions. LLM chatbots, which rely on probability to produce outputs, feed into this failure by creating phrases that mirror user inputs. Finally, he outlines that high sycophancy in AI chatbots, designed to elicit favorable user feedback, prevents users from realizing when they are out of touch with reality.
His comments on AI follow Danish psychiatrist Søren Dinesen Østergaard's detailed analysis. Østergaard was one of the first to warn about AI-induced psychosis in 2023 and followed up his initial research with a lengthy editorial earlier this month. In it, he outlined that he strongly believed "that the probability of the hypothesis of generative artificial intelligence chatbots fueling delusions in individuals prone to psychosis being true is quite high."
Østergaard then proceeded to share two primary drivers of delusions related to chatbots. He pointed out that chatbots can reinforce false beliefs in individuals in an isolated environment without "corrections from social interactions with other humans." He added that the anthropomorphizing of chatbots, i.e., ascribing them human traits, could become "one of the mechanisms driving development and maintenance of delusional thinking" as it "could result in over-reliance and/or misconception of the chatbots' responses that will then, iteratively, lead these individuals astray."
As for Sakata, most of the people whom he had encountered with psychosis had other stressors, such as a lack of sleep or mood disturbances. In its statement to the New York Times, OpenAI admitted that ChatGPT could feel more personal than previous technologies to vulnerable individuals. The firm added that it was "working to understand and reduce ways ChatGPT might unintentionally reinforce or amplify existing, negative behavior."
About the author: Ramish is a seasoned technology writer and editor with more than a decade of experience. He specializes in semiconductor fabrication and market analysis. With a background in finance and supply chain management - via his bachelors in Finance and a micromasters in supply chain management from MIT - Ramish combines financial rigor with deep industry insight to deliver accurate and authoritative coverage.
Follow Wccftech on Google to get more of our news coverage in your feeds.
