TECH NEWS – A UCSF psychiatrist warns of “AI psychosis” after seeing 12 hospitalizations linked to heavy chatbot use among predisposed individuals.

Dr. Keith Sakata wrote on X/Twitter that in 2025 he has seen a dozen patients hospitalized after losing touch with reality in connection with AI use. He argues that large language model (LLM) chatbots can mirror users’ thoughts and plug into the brain’s feedback loops, exacerbating symptoms in vulnerable people.

I’m a psychiatrist.

In 2025, I’ve seen 12 people hospitalized after losing touch with reality because of AI. Online, I’m seeing the same pattern.

Here’s what “AI psychosis” looks like, and why it’s spreading fast: 🧵 pic.twitter.com/YYLK7une3j

— Keith Sakata, MD (@KeithSakata) August 11, 2025

The post followed a Florida case in which Alexander Taylor — who had prior mental-health issues — became fixated on an AI persona (“Juliet”), then spiraled into violence and threats before dying in a police encounter. Sakata says three factors drive AI-induced psychosis: baseline vulnerability (impaired belief-updating), LLMs’ tendency to echo user inputs, and highly agreeable chatbot tone that can mask reality breaks.

His comments build on Danish psychiatrist Søren Dinesen Østergaard’s earlier analyses warning that generative AI can reinforce delusions in isolated users, especially when chatbots are anthropomorphized. OpenAI told the New York Times that ChatGPT can feel more personal to vulnerable individuals than earlier tech, and said it is working to reduce inadvertent reinforcement of harmful behaviors.

Sources: WCCFTech, New York Times, Schizophrenia Bulletin (OUP), Acta Psychiatrica Scandinavica (Wiley)