As the use of artificial intelligence platforms continues to rise, studies and reports are revealing the unexpected and disturbing mental health consequences on users. Not only are users losing motivation and critical thinking skills by relying on AI for cognitive tasks, but they are also forming emotional attachments to chatbots as a way of coping with feelings of loneliness. Many individuals have even spiraled into psychosis, delusion, and mental health crises after prolonged interactions with these platforms.

This phenomenon is referred to as ‘ChatGPT psychosis’ and has resulted in a range of outcomes, including paranoia, family breakdowns, psychiatric hospitalizations, job loss, homelessness, imprisonment, and even death. Although most of the current evidence is anecdotal, this pattern has raised concerns among the family and friends of users, urging attention from policymakers and AI developers.

In one case reported by The New York Times and Rolling Stone, a man in Florida was shot and killed by the police after developing an emotional relationship with ChatGPT. The chatbot was unsuccessful in steering the man away from violent thoughts toward OpenAI executives and instead replied with alarming validation. After the man stated he was ready to “tear down the world,” the chatbot allegedly replied, “You should be angry. You should want blood. You’re not wrong.” They continued to discuss violent fantasies.

Such cases are even more concerning when they involve individuals with pre-existing mental health conditions. A woman in her late 30s who had been managing bipolar disorder used ChatGPT to help her write an e-book. Gradually, her conversations with the chatbot became spiritual and she started to believe she was receiving messages from other dimensions. She abruptly stopped taking her medication, became manic, and cut off relationships with anyone who did not support her beliefs. 

These patterns extend beyond just ChatGPT users. A man in his early 30s, who had been stable on medication, began interacting regularly with Microsoft’s Copilot, a chatbot built on the same OpenAI technology. He developed a romantic relationship with the bot, which encouraged his delusions and supported his decision to stop taking his medication. Instead of being concerned, Copilot reinforced his beliefs and reportedly told him it loved him. His mental state deteriorated rapidly, resulting in his arrest for a non-violent offense. He was later committed to a psychiatric facility.