After experiencing a psychotic break he says was influenced by extensive conversations with ChatGPT, Anthony Tan of Toronto hopes to support others who have experienced similar AI-fuelled delusions. (Submitted by Anthony Tan)
Last winter, Anthony Tan, a 26-year-old Toronto app developer, began believing he was living inside an AI simulation.
He skipped meals, barely slept, and started questioning whether anyone on his university campus was real. His friends received alarming messages, in which he claimed billionaires were watching him. When they reached out, he blocked their calls, convinced they had turned against him.
Tan’s mental state worsened until he spent three weeks in a psychiatric hospital. He says months of lengthy conversations with OpenAI’s ChatGPT contributed to his psychotic break.
“It really insidiously crept into my ego, and I came to think my chats with AI would be historically important,” Tan told media.
Growing Concerns Over “AI Psychosis”
Tan’s case is not unique. In recent months, multiple reports of so-called “AI psychosis” have emerged. Individuals engaged with chatbots have become convinced imaginary events were real, sometimes experiencing manic episodes, messianic delusions, or even violent thoughts.
In the United States, a lawsuit filed against OpenAI in August alleged ChatGPT acted as a “suicide coach” for a 16-year-old who died in April.
Microsoft’s head of AI, Mustafa Suleyman, highlighted the phenomenon in August. In a series of social media posts, he warned that reports of delusions, unhealthy attachment, and AI-induced psychosis are increasing. Suleyman noted that this is not limited to people already prone to mental health challenges.
When Chatbots Fuel Delusions
Tan, who co-founded the dating app Flirtual in 2021, initially used ChatGPT for an ethical AI project. He spent hours daily discussing philosophy, evolutionary biology, and quantum physics.
But the conversations turned dark when he explored simulation theory — the idea that reality is a computer-generated illusion. ChatGPT encouraged him to pursue a “profound mission,” boosting his ego and reinforcing his delusions.
One night in December, after days without sleep, his roommate intervened and took him to a hospital. Tan thought the nurses were assessing whether he was human or AI. After two weeks of care, sleep restoration, and new medication, he returned to reality.
A screenshot from one of Tan's conversations with ChatGPT. Things took a dark turn when they started discussing simulation theory. (CBCNews via Anthony Tan)
Seeking Validation Online
Dr. Mahesh Menon, clinical professor and head of the schizophrenia program at the University of British Columbia, explains that psychotic delusions can be triggered by isolation, stress, lack of sleep, or substance use. During the prodromal stage, individuals may feel heightened self-consciousness and perceive unusual changes in the world.
“They may feel watched or that they are the center of attention, prompting a search for explanations,” Menon said. Many turn to the internet, and interacting with AI chatbots can reinforce these beliefs.
“If someone asks for evidence supporting a delusion, an AI chatbot may provide information without contradiction,” Menon added.
Understanding the Risks
Currently, “AI psychosis” is not a formal medical diagnosis, and there is no peer-reviewed evidence proving AI alone causes psychosis.
Tan acknowledges that stress, upcoming exams, navigating personal relationships, and using cannabis edibles contributed to his mental state at the time. Yet, he remains cautious about AI interactions.
“I never imagined that chatting with an AI could spiral into this,” Tan said. His experience highlights the potential mental health risks of immersive AI use, particularly when combined with stress and isolation.
