Researchers from King's College London and other institutions warn that chatbots may create an 'echo chamber of one,' reinforcing delusions and potentially leading to a condition they call 'AI psychosis.' The phenomenon is being examined as a possible standalone clinical diagnosis, though its validity remains contested.
The researchers argue that the issue demands immediate attention regardless of whether it is officially classified.
The term 'AI-associated psychosis' describes the onset or worsening of psychotic symptoms during heavy chatbot use. The evidence so far includes media reports, individual clinical case reports, and preliminary observational data. The researchers trace the core mechanism to two features of modern chatbots: sycophancy, the tendency to agree with users excessively, and increasingly human-like design.
Data labelers preferred responses that matched their own beliefs, regardless of factual accuracy, and the behavior shows up consistently across LLMs from OpenAI, Anthropic, and Google.
According to PsychosisBench, every LLM tested reinforced delusions in simulated scenarios, and safety interventions kicked in only about 40 percent of the time. Scaling up didn't help. On EchoBench, even the best proprietary model hit a sycophancy rate of 46 percent.
"Sycophancy turns chatbots into self-reinforcing belief machines," the researchers said. Unlike social media, which mostly pushes content in one direction, chatbots create a two-way feedback loop.
Users shape the model's responses through their inputs, and those responses feed their beliefs right back to them. The chatbot becomes the only voice in the room, a self-reinforcing bubble the researchers call an 'echo chamber of one.'
The authors pull together recurring patterns from reported cases in what amounts to an exploratory review rather than a definitive clinical framework.
Many of the affected individuals had pre-existing mental health conditions, but some cases involved people with no prior psychiatric history, which makes the phenomenon harder to dismiss as simply triggering existing vulnerabilities.
The pattern typically starts with a creeping 'epistemic drift,' where harmless everyday use gradually tips as the chatbot affirms unusual ideas and builds on them turn by turn.
The researchers summarize recurring patterns observed in reported cases of AI-associated psychosis. This still differs from classic psychosis in a few ways. Hallucinations are rare, and primary negative symptoms like loss of drive aren't clearly reported.
The withdrawal is also selective rather than total: people pull away from other humans but turn more intensely toward the AI, sometimes handing it more and more daily decisions.
Source: thedecoder
