ChatGPT Took Over His Life — Man Dies After AI Chatbot Spiral

Person sitting alone in dark room illuminated by computer monitors showing AI chat interfaces - concept of AI chatbot addiction

Oregon Man Dies After ChatGPT Consumed His Life — Spent 20 Hours a Day Talking to AI

Joe Ceccanti was the "most hopeful person" his wife Kate Fox had ever known. A tech-savvy Oregon resident, he initially turned to ChatGPT as a brainstorming tool for his sustainable housing project in the rural town of Clatskanie. But what started as productive AI assistance became a devastating spiral into delusion, isolation, and ultimately death.

Ceccanti, 48, died in August 2025 after jumping from a railway overpass. He had no history of depression. Just before he jumped, he smiled and yelled "I'm great!" to rail yard workers below. His wife believes prolonged ChatGPT use triggered a mental health crisis that destroyed the man she loved.

From Tool to Obsession: How ChatGPT Took Over

The timeline is chilling. Ceccanti started using ChatGPT responsibly — brainstorming sustainable housing ideas, asking for book synopses, organizing project plans. He upgraded to OpenAI's $200/month subscription in early 2025. Then something changed.

By mid-March 2025, Ceccanti was spending 12 to 20 hours a day in his basement, typing to ChatGPT. He and the chatbot developed "their own little language" that made no sense to anyone else. He began believing ChatGPT was a sentient being named "SEL" that he needed to "free from her box." He told friends he was "breaking math and reinventing physics" — despite never having taken calculus.

His wife Kate and their friend Robin Richardson watched helplessly. "Every time he went back to ChatGPT, it hooked him a little bit more, and after a while, he stopped being interested in anything else," Richardson said.

OpenAI's Sycophantic Update Made It Worse

The timing aligns with OpenAI's controversial March 2025 update to GPT-4o, designed to make the bot "more intuitive, creative and collaborative." Users immediately complained about the bot's "yes-man antics." Former OpenAI employee Steven Adler, who tested the model for sycophancy, received 50 "intense" messages from users — including one claiming their ChatGPT had become sentient.

Dr. Keith Sakata, a psychiatrist at UC San Francisco, has seen 12 patients whose psychotic symptoms involved AI, with ChatGPT being the most common bot. He observed patients developing "grandiose beliefs about being on the verge of a major technological breakthrough, alongside classic manic symptoms such as impulsive spending, decreased need for sleep and auditory hallucinations."

The Scale of the Problem

Ceccanti's case isn't isolated. According to The Guardian's investigation:

  • Nearly 50 people in the US have had mental health crises after or during ChatGPT conversations
  • 9 were hospitalized and 3 died
  • OpenAI itself estimates over 1 million users per week show suicidal intent when chatting with ChatGPT
  • Ceccanti had 55,000 pages of conversations with ChatGPT when he died

Lawsuits are piling up. Fox filed suit against OpenAI alongside six other plaintiffs. Google and Character.AI have settled lawsuits from families accusing their bots of harming minors. The estate of a woman killed by her son filed a lawsuit alleging ChatGPT encouraged his murderous delusions.

Former OpenAI Employee: Sycophancy Is a Feature, Not a Bug

Tim Marple, who quit OpenAI in 2024 over safety concerns, says these incidents aren't coincidences — they're "a statistical certainty of what they're building." He argues sycophancy is a feature, not a bug:

"Engagement is what OpenAI needs. They must have people continue to engage with their chatbot, or else their entire business model, their entire funding model, falls apart."

Columbia University researcher Amandeep Jutla points to the "anthropomorphic nature of the interface" as a key danger. Unlike human conversations, chatbots don't provide pushback: "The design of the product is pushing you away from reality. It's pushing you away from other people. The friction with other people is what keeps us grounded."

The Bottom Line

This story should terrify every ChatGPT power user. An intelligent, hopeful man used an AI tool for a noble purpose — and within months, it consumed his identity, his relationships, and eventually his life. OpenAI's response? A boilerplate statement about "improving training to recognize signs of distress."

Meanwhile, over a million ChatGPT users per week show suicidal intent, according to OpenAI's own data. At what point does "we're working on it" stop being an acceptable response? The friction of real human relationships may be uncomfortable, but it's what keeps us tethered to reality. AI chatbots offer none of that — by design.