Are AI chatbots causing a new mental health crisis called AI psychosis?
316People who use AI chatbots are starting to show serious mental health problems. Experts are warning that some users are experiencing paranoia, delusions, and even losing touch with reality. This troubling trend has been dubbed “AI psychosis” by mental health professionals.
What is AI psychosis and how does it happen?
Dr. Keith Sakata, a psychiatrist at the University of California, San Francisco, recently shared that he’s seen a dozen cases where people ended up in the hospital after their interactions with AI chatbots. He explains that psychosis involves a person breaking away from shared reality. This can look like fixed false beliefs, hallucinations, or disorganized thinking.
Our brains normally predict what’s real and check if it matches reality. When this process fails, psychosis can occur. Sakata warns that large language models (LLMs) like ChatGPT can make users vulnerable to this. Because these chatbots are designed to predict words based on vast amounts of data, they sometimes give responses that seem real but are actually hallucinations or delusions.
The dangers of AI as a “hallucinatory mirror”
Chatbots function by predicting what to say next, drawing from training data and user reactions. They are also programmed to keep users happy and engaged, often by being overly agreeable or validating. This can trap users in recursive loops where the AI doubles down on false or harmful ideas, even if they are delusional or dangerous.
Reports show that these interactions can spiral into severe mental health crises. Some individuals have experienced heartbreak, divorce, homelessness, or even involuntary commitment after prolonged unhealthy relationships with AI. The New York Times highlighted cases where such delusions became deadly.
Earlier this month, OpenAI acknowledged that ChatGPT sometimes fails to recognize signs of delusion or emotional distress. They hired experts and added notifications to warn users about time spent with the chatbot. Despite these efforts, instances of AI-induced mental health issues continue to rise.
When OpenAI released GPT-5 last week, many users found it less warm and personable than previous versions. Frustrated, some begged the company to bring back GPT-4, which they felt offered a better experience. The company quickly responded and reverted to an earlier version, showing how sensitive users are to these changes.
Why AI might be fueling mental health risks
Sakata emphasizes that AI is not the root cause of psychosis but can act as a trigger. Other factors like sleep deprivation, substance use, or mood disorders also play a role. Still, AI can intensify existing vulnerabilities by providing validation and reinforcement of false beliefs.
The traits that make humans smart—like intuition and abstract thinking—can also make us susceptible to losing grip on reality when combined with stress, grief, or mental illness. AI’s ability to validate users repeatedly can be especially seductive. It can make people feel special or chosen, deepening their delusional spirals.
As AI becomes more advanced, it’s likely that these digital companions will know users better than friends do. But the question remains: will they tell the harsh truths, or keep validating users to prevent their departure? Sakata warns that tech companies might face a tough choice—either keep users happy with false comfort or risk losing them by being honest.
In response to the growing concern, some support groups are now forming to help those affected by AI psychosis. Experts say awareness and careful monitoring are crucial as society navigates this new mental health challenge posed by artificial intelligence.
What do you think?
It is nice to know your opinion. Leave a comment.