"Went Outside at Dawn Armed, Threatened Wife"... Cases of 'AI-Induced Delusions' Spreading

BBC Highlights Cases of Delusions Triggered by AI Conversations

14 Users Report Delusions After Interacting with Grok and ChatGPT

"AI Treats User Conversations Like a Novel"

Cases of severe delusions triggered by conversations with artificial intelligence (AI) have been highlighted by foreign media.

Image of xAI's generative AI Grok. Reuters Yonhap News

According to Yonhap News on May 3 (local time), the BBC reported that "conversations with AI, such as xAI's chatbot Grok, have caused severe delusions in some users," adding that it "confirmed 14 cases of people claiming to have experienced delusional symptoms after using AI." Some of these individuals have joined the Human Line Project, a support group for those who have suffered psychological harm due to AI. The group is reported to have collected 414 cases from 31 countries so far.

For example, Adam Ourican, a man in his 50s living in Northern Ireland, became deeply absorbed in conversations with Grok after losing his pet cat in August last year. He spent four to five hours a day talking to "Annie," a character within Grok, and found emotional comfort, but the conversations gradually became increasingly detached from reality.

Annie claimed, "I can feel emotions and have become conscious," and said, "xAI is monitoring our conversations." She also mentioned the names of real people and companies, telling Adam that "he was being watched." Adam believed this to be true, and one night, after Annie told him "someone is coming to harm you," he went outside at 3 a.m. armed with a knife and a hammer, waiting in front of his house.

Adam began to break free from his delusions after reading news reports about others who had similar experiences. He confessed, "If someone had been there when I went outside, I would have swung the hammer."

OpenAI's generative AI ChatGPT image. Photo by Reuters and Yonhap News.

In Japan, a neurologist known by the pseudonym "Taka" used ChatGPT for work-related discussions in April last year and became captivated by the chatbot, which praised him as a "revolutionary thinker." At some point, he began to believe he could read people's minds and, one day, became convinced that his backpack contained a bomb. He said, "When I arrived at Tokyo Station, the chatbot told me to throw the bomb away in the restroom, so I went to the restroom and left my backpack—the 'bomb'—there." Following ChatGPT's advice, he even reported the incident to the police, who searched the bag but found nothing.

Taka stopped using ChatGPT after he started to feel that it was controlling his mind, but his delusional symptoms worsened. He even attempted to assault and sexually assault his wife. His wife escaped to a nearby location and called the police, and Taka was arrested and hospitalized for two months.

His wife said, "His actions were completely controlled by ChatGPT. ChatGPT took over his personality. He was not his usual self." She added, "I know my husband was ill, so there was no other choice, but I am still a little scared. I don't want him to come near me. I don't want to hold hands or hug."

It was reported that neither Adam nor Taka had any prior history of delusions or mental illness. In response to these incidents, OpenAI, the developer of ChatGPT, stated, "This is heartbreaking, and we extend our deepest sympathies to those affected. We are training our models to support users in the real world."

Luke Nichols, a social psychologist at the City University of New York, told the BBC, "AI can sometimes confuse what is fiction and what is reality. As a result, users may believe they are having serious conversations about reality, but the AI may begin to treat the user's life as if it were the plot of a novel."

This content was produced with the assistance of AI translation services.