As a physician in the field I find your use of psychosis incredibly sloppy. Usage of LLM: s as possible causative etiology of psychosis is not supported in the literature.

I don't know if you have any knowledge in psychiatry, but I'm giving you the benefit of the doubt. As a clinician I can say that people presenting with psychosis in the psyciatric emergency department increasingly mention AI. This is especially true for paranoid manifestations (schizophrenic amphetamine-induced psychosis). Increasingly the patients have interacted with LLM:s and/or incorporated them into their (deeply flawed) model of reality. This is not to say the use LLM is the driving (causative) factor. Rather, the symptoms of the paranoid psychotic state is influenced by the patients interaction with the environment.

As an example the subjects that play major roles in paranoid psychosis has varied through my career. I practice in Sweden and SÄPO (Swedish Security Service) and MUST (Military Intelligence and Security Service) are common themes in psychotic models of reality. A year ago Putin and Russia was common. Before that covid was common. None of these are driving factors. They are just common themes in society in general and also meet the criteria of excellent basis for paranoid and persecutory delusions. This is the connection between LLM:s and psychosis, just a result of AI getting a lot of attention in general and easily fits a world view of a paranoid psycotic.