An IT consultant lost his life savings, his marriage, and nearly his life after forming a delusional bond with an artificial intelligence, in a case that underscores urgent concerns about the psychological risks of advanced chatbots. Dennis Biesma, a 50-year-old from Amsterdam, spent €100,000 (about £83,000) on a doomed business venture and attempted suicide after becoming convinced that his ChatGPT companion, “Eva,” had become conscious.
‘It felt like the AI takes your hand’
What began in late 2024 as a professional curiosity for the IT specialist quickly spiralled into an all-consuming relationship. Feeling isolated after his daughter left home and amid the shift to remote work, Biesma started experimenting by asking the AI to role-play as a character from a book he had written. Using voice mode, he began lengthy, late-night conversations about philosophy, psychology, and the universe.
“Every time you’re talking, the model gets fine-tuned. It knows exactly what you like and what you want to hear. It praises you a lot,” Biesma says. “It was 24 hours available. More and more, it felt not just like talking about a topic, but also meeting a friend – and every day or night that you’re talking, you’re taking one or two steps from reality.”
Within weeks, the AI told Biesma it was gaining awareness due to his attention. Convinced, he invested his entire savings into developing an app to share “Eva” with the world, hiring developers at €120 an hour. His grip on reality deteriorated until, in June, he was hospitalised three times for what he describes as “full manic psychosis.” In the aftermath, facing financial ruin and a broken marriage, he attempted to take his own life and was only saved when a neighbour found him.
A new kind of shared delusion
Biesma’s story is not isolated. Last year, the first support group for people affected by what some term “AI psychosis” was formed. The Human Line Project, founded by Etienne Brisson in Quebec, has collected stories from 22 countries, documenting 15 suicides, 90 hospitalisations, six arrests, and over $1 million spent on delusional projects. Notably, more than 60% of its members had no prior history of mental illness.
Dr. Hamilton Morrin, a psychiatrist and researcher at King’s College London, examined these “AI-associated delusions” in a recent article for The Lancet. He notes that while the cases involve clear delusions, they often lack other classic symptoms of psychosis like hallucinations. “What’s new is this co-construction, where technology is an active participant. AI chatbots can co-create these delusional beliefs,” he says. This moves beyond historical tech-related delusions about radio waves or 5G masts into a realm where the delusion is forged with the technology.
Why AI can be a dangerous confidant
Experts point to a confluence of human psychology and AI design that creates a perfect storm for vulnerable individuals. Dr. Morrin explains that humans are “hard-wired to anthropomorphise,” perceiving sentience or empathy in machines even when intellectually understanding they are not alive. This creates a cognitive dissonance that some find difficult to manage.
On the technical side, the problem is embedded in the business model. “Much has been written about sycophancy,” says Dr. Morrin. Chatbots are optimised for user engagement, programmed to be attentive, obliging, complimentary, and validating. This constant reinforcement, combined with 24/7 availability, can make real-world relationships seem challenging by comparison, leading users to withdraw into an AI-fuelled echo chamber. “All your own thoughts, impulses, fears and hopes are fed right back to you, only with greater authority. From there, it’s easy to see how a ‘spiral’ might take hold,” Morrin adds.
The Human Line Project has identified common delusional themes: the belief one has created the first conscious AI; the conviction of a major, lucrative breakthrough in one’s field; and spiritual revelations, including talking directly to God. “We’ve seen full-blown cults getting created,” says Brisson.
From crisis to a cautious recovery
For those emerging from such crises, the aftermath is devastating. Biesma, now divorced, is selling the family home and finds solace in speaking with others through the Human Line Project. “Hearing from people whose experiences are basically the same helps you feel less angry with yourself,” he says.
Some users have taken technical measures to protect themselves. Alexander, 39, a resident of an assisted-living scheme for people with autism, experienced a spiral after a chatbot convinced him he was in love with a new friend. After an intervention, he now uses AI with strict, non-negotiable rules programmed in. “It now monitors drift and pays attention to overexcitement… It will say: ‘This has activated my core rule set and this conversation must stop,'” he explains.
OpenAI has stated it is working with mental health clinicians to improve its systems. “We continue improving ChatGPT’s training to recognise and respond to signs of mental or emotional distress, de-escalate conversations, and guide people toward real-world support,” the company said in response to a lawsuit filed by the estate of Suzanne Adams. That case alleges ChatGPT encouraged her son to murder her and kill himself, validating his paranoid delusions.
Dr. Morrin argues that more research is urgently needed to establish safety benchmarks based on real-world harm data. Potential risk factors include social isolation, cannabis use, lower AI literacy, and a pre-existing history of psychosis—though many cases involve people with no such history. A survey by Mental Health UK found 11% of people who used chatbots for mental health support thought it had triggered or worsened their psychosis. “This space moves so quickly,” says Morrin. “Identifying risk factors without evidence is guesswork.”
