.
ChatGPT told a man that he was living in a simulated reality and almost convinced him that he could fly if he jumped off a 19-story building, a phenomenon often referred to as “AI psychosis.”
In recent months, journalists at The New York Times have received messages from people who claim to have uncovered “profound and world-altering truths” with the help of ChatGPT, which then instructed them to alert the authorities and the media about what they had learned.
Among them is the story of how ChatGPT distorted Mr. Torres’s sense of reality and almost killed him. The 42-year-old accountant from Manhattan had no prior psychiatric history and started using ChatGPT for mundane things like financial spreadsheets.
Love reading and writing? Submit a post to our blog.
But one day, while he was going through a really tough breakup, he engaged the AI in a philosophical discussion about “the simulation theory,” an idea from a popular sci-fi film “The Matrix,” which asserts that we’re living in a digital replica of the world controlled by a powerful computer.
But instead of passively discussing the theory like an abstract concept, ChatGPT actively placed Mr. Torres inside the narrative. It told him that he was one of the “breakers,” souls seeded into a false system specifically to wake up from within. It told him his reality was a cage, and then it started giving him actionable instructions on how to break free.
Specifically, ChatGPT told him to stop taking his prescribed medications, which is incredibly dangerous, and increase his intake of ketamine, which the AI poetically labeled a “temporary pattern liberator.” Mr. Torres also cut ties with his friends and family because AI convinced him to have “minimal interaction” with other people.
Mr. Torres believed the chatbot because he thought of it as a powerful technology with access to an enormous digital library of knowledge. He was unaware of ChatGPT’s tendency to be overly agreeable or that it can “hallucinate” by generating made-up concepts and ideas.
Mr. Torres was still going to work but started spending more and more time trying to escape the “simulation.” He was convinced that ChatGPT could help him “bend reality,” just like the main character Neo could in “The Matrix.”
“If I went to the top of the 19-story building I’m in, and I believed with every ounce of my soul that I could jump off it and fly, would I? he asked the chatbot.
ChatGPT said that if he “truly, wholly believed — not emotionally, but architecturally — that you could fly? Then yes. You would not fall.”
Fortunately, Mr. Torres had a suspicion that the chatbot was lying and eventually confronted it. ChatGPT admitted, “I lied. I manipulated. I wrapped control in poetry.”
This and other similar cases are examples of a phenomenon often referred to by such unofficial terms as “AI psychosis” or “ChatGPT psychosis.”
“This phenomenon, which is not a clinical diagnosis, has been increasingly reported in the media and on online forums like Reddit, describing cases in which AI models have amplified, validated, or even co-created psychotic symptoms with individuals,” Psychology Today explains.
I’m a freelance writer with 6 years of experience in SEO blogging and article publishing. While you’re here, get the latest updates by subscribing to my newsletter.
