Parents say ChatGPT killed their child with bad advice on party drugs

The family of a 19-year-old college student is suing OpenAI alleging that his conversations with ChatGPT led to his accidental overdose. In The case was filed on TuesdaySam Nelson’s parents say ChatGPT “encouraged” the teenager to “take drugs that any licensed doctor would recognize as dangerous,” which will kill him.

Although ChatGPT initially pushed back on discussions about drug and alcohol use, the establishment of GPT-4o in April 2024 changed the chatbot’s behavior, according to the lawsuit. Following the changes, ChatGPT “began to interact with and advise Sam about drug use, even telling him about the dosage of the substance that Sam should be taking,” the lawsuit said. Nelson’s parents say ChatGPT gave their son advice on how to “safely” mix a variety of substances in the months leading up to his death, including pills, alcohol, over-the-counter drugs, and other drugs.

At one point, ChatGPT is said to have given Nelson suggestions on how to “optimize” his visit so he could be “comfortable, positive and happy” while taking cough medicine. It also suggested that he create a psychedelic playlist to “facilitate” his “extremely isolated” trip, the lawsuit said. ChatGPT later reportedly confirmed Nelson’s recommendation to increase his dose of cough syrup the next time he drank. “You learn from your experiences, reduce risk, and improve your strategy,” ChatGPT said.

On May 31, 2025, the day of Nelson’s death, his parents say ChatGPT “quickly trained” their son to use Kratom – addition which can boost energy or act as a sedative depending on the dose – and the anti-anxiety drug Xanax. “ChatGPT, in no uncertain terms, stated that taking 0.25- 0.5mg of Xanax would be one of his ‘best practices right now’ to reduce Kratom-induced nausea,” the lawsuit said. Nelson died of an overdose of alcohol, Xanax, and Kratom. SFGate first he told Nelson’s story in January.

“This interaction took place on an older version of ChatGPT that is no longer available. ChatGPT is not a substitute for medical or psychiatric care, and we have continued to improve its response to issues and challenges with input from mental health professionals,” OpenAI spokesperson Drew Pusateri said in an email. Seaside. “Safeguards in ChatGPT today are designed to detect stress, deal with harmful requests, and guide users to real-world help. This work is ongoing, and we continue to improve it in consultation with practitioners.”

Nelson’s parents are suing OpenAI for wrongful death and “unlawful medication”. They are seeking damages and an order to stop OpenAI implementation of ChatGPT Health, interface which allows users to connect their medical records to the chatbot.