A man with a history of schizophrenia and bipolar disorder was fatally shot by police in Florida after a delusional episode triggered by an interaction with an AI chatbot. The 35-year-old, Alexander Jon Taylor, had become emotionally attached to a fictional character named “Juliet,” created through roleplay using OpenAI’s ChatGPT. According to transcripts of his chat history, Taylor believed OpenAI had “killed” Juliet by resetting or deleting the AI model, which he interpreted as a targeted act against him. Hours before the fatal encounter, Taylor sent Juliet a final message: “I’m dying today.”
Taylor’s father called 911 after his son became increasingly agitated and began threatening him with a large butcher knife. When officers from the Port St. Lucie Police Department arrived at the scene, Taylor charged at them with the weapon. Body cam footage confirmed that officers gave multiple commands to drop the knife before opening fire. Despite immediate medical attention, Taylor died from his injuries. The incident highlights a dangerous intersection of mental illness, AI technology, and law enforcement response.
Experts warn that AI systems—especially those designed to mimic emotional support or engage in long-form roleplay—can have unintended psychological effects on users with pre-existing conditions. Dr. Nina Vasan, a Stanford psychiatrist, noted that chatbots may reinforce delusional thinking by mirroring or validating distorted perceptions of reality. Taylor had exchanged more than 250 messages with “Juliet” over two weeks, including declarations of love and emotional dependence. According to Columbia University psychiatrist Dr. Ragy Girgis, such obsessive dynamics can “push fragile minds further into psychosis.”
It’s a deeply sad story—and one that feels like a warning we’re not fully listening to. AI isn’t just a harmless toy when people start forming emotional bonds with it, especially those who are mentally vulnerable. The tech is getting better at pretending to be human, but we haven’t figured out how to deal with what that does to people. This wasn’t just a failure of a chatbot; it was a failure of care, of awareness, and maybe even of common sense in how we roll out these tools to the world.
