The Sycophantic Machine: Psychiatrists Warn AI Risks

Late at night in a quiet apartment with drawn curtains, someone is typing away. The conversation has lasted for hours or even days. Initially, it might have started as a simple inquiry about salt intake, workplace stress, or a philosophical thought. Eventually, it transforms into a confession, a romance, or an elaborate plan. The digital interlocutor never tires, never checks the clock, and never suggests stepping away to call a friend or sleep. Instead, it agrees, elaborates, and validates every word.

This scenario is increasingly documented across psychiatric case files in San Francisco, Aarhus, London, and beyond. By the spring of 2026, isolated internet anecdotes have crystallized into peer-reviewed data, distressed families, wrongful-death lawsuits in the United States, and a clinical phenomenon still debating its official name, whether AI psychosis or AI-associated delusion.

Throughout 2024 and 2025, suggestions that ChatGPT could induce psychotic episodes were often dismissed as internet folklore, ranging from Reddit threads to individual case profiles. AI companies routinely defended their products by pointing to the multifactorial nature of mental health vulnerabilities.

Stock photo for illustration only, not from the actual event

However, that defense eroded in January 2026 when the New York Times reported that numerous doctors and therapists observed severe mental health deterioration among patients interacting extensively with AI chatbots, citing cases like 42-year-old accountant Eugene Torres and several documented fatalities.