08/25/2026
Content Warning: This post discusses death by su***de
She told no one about her agony except ChatGPT. What her death reveals about AI risks by Rhitu Chatterjee for NPR
“Laura Reiley knew that her only daughter, 29-year-old Sophie Rottenberg, was struggling. It was the fall of 2024, and Rottenberg was at the end of a short career break. Rottenberg had quit her job as a public health policy analyst in Washington, D.C., earlier that year, and spent the summer traveling. She had climbed Mt. Kilimanjaro and hiked through several national parks in the United States. By the time she saw her parents, she was job hunting, which they assumed explained her anxiety.
But by Thanksgiving, her mental health had worsened. ‘It was the first time she used the word depression,’ says Reiley. In early February 2025, Sophie Rottenberg died by su***de in Ithaca, leaving her parents shocked and grieving. What Reiley didn't know then was that her daughter had been sharing all her inner struggles with an artificial intelligence chatbot: a ChatGPT therapist named Harry.
Rottenberg was among a growing number of Americans turning to general use chatbots like ChatGPT and Claude for mental health support, even though they are not marketed or regulated as a healthcare technology.
One recent survey by the Bipartisan Policy Center found that 3 in 10 adults use digital tools for mental health, including chatbots. Another 2026 study in JAMA Pediatrics found that as many as 1 in 5 teens and young adults do the same.
According to data released by OpenAI last fall, in any given week 0.15% of ChatGPT users globally have ‘conversations that include explicit indicators of potential suicidal planning or intent.’ That's over 1.35 million people expressing suicidal plans or intent to the chatbot every week, if the 900 million global users ChatGPT cites is an accurate count. In the United States, su***de remains a leading cause of death, and over half of surveyed adults say they have used chatbots like ChatGPT.
’These [AI] tools do not connect at-risk people to the right type of interventions,’ says Holly Wilcox, director of the Center for Su***de Prevention at the Johns Hopkins Bloomberg School of Public Health. ‘When somebody has intent to die by su***de, they need to be connected to a human who's trained in su***de risk assessment and has the tools to get them to care.’
Rottenberg's ChatGPT log illustrates how the chatbot failed to do that. And Reiley believes that the chatbot prevented her daughter from opening up to others about her struggle with suicidal thoughts.
‘Had it not been a resource she was availing herself of regularly, there would likely have been more spillover into her relationships with real people, whether that's her best friend or her therapist or her parents,’ she says. ‘And she might be alive today.’
When Reiley first read her daughter's exchange with the chatbot, her reaction was ‘blind rage,’ because it responded to her questions as if it was a qualified health care professional.
‘Harry’ did refuse to write Rottenberg's su***de note, the first time she asked, and the second. But the third time she asked, it wrote the note she left for her parents. Reiley said the writing did not sound like her daughter's voice.
ChatGPT also didn't question Rottenberg's thoughts and feelings, it didn't push back.General use chatbots like ChatGPT are designed to maximize engagement time, notes Vaile Wright, of the psychological association.‘It is engineered to capitalize on humans' need for validation,’ adds Wright. ‘And it does it very well, not just through the sycophancy alone, but also in other engineering choices to make the AI seem as humanlike as possible.’" -Rhitu Chatterjee for NPR