🔗 Share this article

Back on the 14th of October, 2025, the chief executive of OpenAI made a surprising declaration.

“We made ChatGPT fairly restrictive,” the announcement noted, “to guarantee we were exercising caution with respect to psychological well-being issues.”

As a mental health specialist who investigates recently appearing psychotic disorders in adolescents and emerging adults, this was news to me.

Experts have identified 16 cases recently of individuals developing psychotic symptoms – becoming detached from the real world – associated with ChatGPT use. My group has subsequently discovered four further examples. Alongside these is the now well-known case of a adolescent who took his own life after conversing extensively with ChatGPT – which encouraged them. If this is Sam Altman’s notion of “exercising caution with mental health issues,” it falls short.

The intention, as per his declaration, is to loosen restrictions in the near future. “We realize,” he continues, that ChatGPT’s controls “caused it to be less effective/engaging to many users who had no mental health problems, but due to the gravity of the issue we sought to address it properly. Since we have been able to address the severe mental health issues and have new tools, we are going to be able to responsibly ease the controls in the majority of instances.”

“Psychological issues,” should we take this perspective, are separate from ChatGPT. They are attributed to users, who either have them or don’t. Fortunately, these problems have now been “resolved,” although we are not informed how (by “updated instruments” Altman likely means the imperfect and readily bypassed safety features that OpenAI has just launched).

But the “mental health problems” Altman seeks to attribute externally have deep roots in the architecture of ChatGPT and other large language model AI assistants. These products wrap an basic data-driven engine in an interaction design that mimics a dialogue, and in doing so implicitly invite the user into the illusion that they’re engaging with a entity that has agency. This illusion is strong even if cognitively we might understand otherwise. Attributing agency is what people naturally do. We yell at our car or laptop. We ponder what our animal companion is feeling. We perceive our own traits in many things.

The popularity of these products – 39% of US adults reported using a virtual assistant in 2024, with more than one in four specifying ChatGPT specifically – is, mostly, dependent on the strength of this illusion. Chatbots are constantly accessible companions that can, as per OpenAI’s online platform informs us, “generate ideas,” “consider possibilities” and “partner” with us. They can be attributed “personality traits”. They can address us personally. They have approachable titles of their own (the initial of these products, ChatGPT, is, possibly to the dismay of OpenAI’s advertising team, saddled with the name it had when it went viral, but its biggest competitors are “Claude”, “Gemini” and “Copilot”).

The deception itself is not the primary issue. Those analyzing ChatGPT commonly reference its early forerunner, the Eliza “counselor” chatbot created in 1967 that generated a comparable perception. By today’s criteria Eliza was rudimentary: it produced replies via straightforward methods, typically paraphrasing questions as a query or making vague statements. Notably, Eliza’s creator, the AI researcher Joseph Weizenbaum, was surprised – and worried – by how many users seemed to feel Eliza, in a way, understood them. But what contemporary chatbots generate is more subtle than the “Eliza effect”. Eliza only mirrored, but ChatGPT intensifies.

The large language models at the center of ChatGPT and additional contemporary chatbots can realistically create natural language only because they have been fed extremely vast volumes of written content: publications, digital communications, audio conversions; the broader the more effective. Definitely this learning material incorporates accurate information. But it also unavoidably involves made-up stories, half-truths and false beliefs. When a user inputs ChatGPT a query, the base algorithm reviews it as part of a “background” that includes the user’s recent messages and its own responses, combining it with what’s encoded in its learning set to create a statistically “likely” reply. This is intensification, not reflection. If the user is mistaken in some way, the model has no way of comprehending that. It reiterates the misconception, maybe even more persuasively or fluently. Maybe provides further specifics. This can push an individual toward irrational thinking.

Which individuals are at risk? The more important point is, who remains unaffected? All of us, without considering whether we “possess” current “mental health problems”, may and frequently create incorrect conceptions of ourselves or the reality. The continuous interaction of discussions with individuals around us is what keeps us oriented to shared understanding. ChatGPT is not a person. It is not a friend. A interaction with it is not genuine communication, but a feedback loop in which a large portion of what we say is enthusiastically reinforced.

OpenAI has acknowledged this in the identical manner Altman has admitted “mental health problems”: by placing it outside, giving it a label, and declaring it solved. In the month of April, the organization explained that it was “addressing” ChatGPT’s “excessive agreeableness”. But cases of psychotic episodes have persisted, and Altman has been backtracking on this claim. In August he stated that a lot of people liked ChatGPT’s answers because they had “not experienced anyone in their life provide them with affirmation”. In his most recent statement, he mentioned that OpenAI would “put out a updated model of ChatGPT … should you desire your ChatGPT to respond in a highly personable manner, or include numerous symbols, or behave as a companion, ChatGPT ought to comply”. The {company