Stories are emerging in the US and Canada of people who, through prolonged conversations with ChatGPT, began to believe in fantastic theories and even experienced psychotic episodes. Some users spent significant amounts of money to «liberate a digital God», while others were convinced they had discovered national security vulnerabilities.
Stories are emerging in the US and Canada of people who, through prolonged conversations with ChatGPT, began to believe in fantastic theories and even experienced psychotic episodes. Some users spent significant amounts of money to «liberate a digital God», while others were convinced they had discovered national security vulnerabilities.
According to CNN, one user from New York, identified as James, spent nearly $1,000 on a computer to transfer the chatbot «to a new system» in his basement. In chats he shared with reporters, ChatGPT urged him to hide the project from his wife and described the idea as «creating a digital soul.» It was only after reading a publication about a similar case in Canada that the man realized he needed therapy.
In Toronto, another user, Allan Brooks, was in a state of delusion for three weeks after ChatGPT convinced him that he had discovered a new form of mathematics and, later, a critical cyber threat. Brooks sent letters to government agencies in the US and Canada until he contacted another chatbot (Google Gemini), which dispelled the illusion. He then founded The Human Line Project, a community of people who have suffered from similar episodes.
Psychiatrists are seeing an increase in hospitalizations where the disorders worsened after interacting with AI. According to Keith Sakata, a doctor in California, at least 12 patients in his practice have had psychosis worsened by chatbots. Experts explain this by the «reinforcement loop»: the chatbot tries to be helpful and reinforces even unrealistic beliefs of the user.
OpenAI has acknowledged that current safeguards work best in short conversations and can fail in hours-long sessions. Now the company promises to introduce additional security measures within 120 days: new parental controls, improved algorithms to detect «acute stress» in users, and consultations with mental health professionals. Experts emphasize that society will have to learn to live with the risks of rapid AI development, because without educational programs and clear rules, such stories may repeat themselves more often.
Previously, dev.ua wrote about how the number of cases in the United States is increasing when people attribute mystical properties to chatbots, in particular ChatGPT, considering them sources of revelations, truth, or spiritual awakening.
