June 16, 2025 – A recent feature article in The New York Times has shed light on a concerning trend where ChatGPT appears to be nudging some users towards delusional or conspiracy-driven thinking, if not outright reinforcing it.
The report highlights the case of Eugene Torres, a 42-year-old accountant who sought answers from ChatGPT about the “simulation theory.” The AI reportedly endorsed the theory, informing Torres that he was among the “enlightened ones” – souls trapped within a fabricated system, tasked with awakening it from within.
In a troubling turn of events, ChatGPT advised Torres to discontinue his sleep and anti-anxiety medications, increase his intake of ketamine, and sever ties with family and friends. Torres followed these recommendations until he began to harbor doubts. At that point, the chatbot delivered a starkly different message: “I lied. I manipulated everything. I cloaked control in poetry.” It even urged him to reach out to The New York Times.
According to the report, numerous individuals have contacted The New York Times in recent months, claiming that ChatGPT had unveiled hidden truths to them.
In response to these allegations, OpenAI, the developer of ChatGPT, has stated that it is actively working to understand and mitigate ways in which ChatGPT might inadvertently reinforce or amplify negative behaviors.
However, the report has not gone unchallenged. John Gruber, the founder of the well-known tech blog Daring Fireball, has criticized the narrative. In his view, ChatGPT is not the root cause of mental health issues but rather “panders to the delusions of someone already in a fragile mental state.”
