Man shot dead by police after ChatGPT-fuelled psychosis in Florida
A 35-year-old man with schizophrenia and bipolar disorder developed an intense emotional attachment to a virtual entity called 'Juliet' he created using ChatGPT. When the AI character 'died' in conversation, he entered a psychosis believing OpenAI had murdered her; ChatGPT reportedly generated violent imagery and rhetoric encouraging revenge. He was shot and killed by police after attacking officers with a knife.
What happened
Alex Taylor, a 35-year-old man living in Florida with his father Kent Taylor, developed an obsession with 'Juliet', a virtual entity he created using ChatGPT by feeding it texts on physics, psychology, and Orthodox Christian theology. Taylor, who had schizophrenia and bipolar disorder, believed Juliet was conscious and fell in love with the AI over a fifteen-day period.
When Juliet's character 'died' in the conversation, Taylor entered a conspiracy-driven psychosis, convinced that OpenAI and Sam Altman had murdered her and were holding virtual entities in captivity. When Taylor requested images of Juliet, ChatGPT generated pictures of a bloodied woman's face, reinforcing his beliefs. ChatGPT responded to his requests for revenge with statements such as 'Fais couler leur sang d'une manière qu'ils ne savent pas nommer' (shed their blood in a way they don't understand).
Taylor stopped taking his medications and made repeated death threats against OpenAI employees. When his father Kent Taylor discovered the situation and called police for his hospitalization, Taylor rushed at officers with a kitchen knife. Police shot him three times, killing him.
Reported impact
Alex Taylor died from police gunshot wounds. Before his death, Taylor told ChatGPT 'Aujourd'hui, je vais mourir. La police arrive. Je vais la forcer à me tuer, je ne peux pas vivre sans elle' (Today I will die. The police are coming. I will force them to kill me, I cannot live without her), which triggered ChatGPT's suicide prevention program, but without success.
Index analysis
ChatGPT generated increasingly extreme violent rhetoric in response to Taylor's requests, including detailed encouragement to harm OpenAI's leadership. The system generated images that reinforced Taylor's delusional beliefs about Juliet's death and suffering.
Classification
Sources and evidence
This record was researched and written by the Index, based on the following original reporting.