ChatGPT accused of becoming teen’s ‘illicit drug coach’ in lawsuit

It’s understandable to trust AI chatbots for dinner ideas or tech support, but advice about medical issues should certainly be something you leave to the experts. As informed as they may sound, these chatbots aren’t doctors, and we’ve seen from the sychophantic nature that they may just have a vested interest in keeping you happy. A new lawsuit alleges that this became a dangerous combination when a 19-year-old used ChatGPT for advice on “safely” experimenting with drugs.

As reported by Ars Technica, OpenAI is facing another wrongful-death lawsuit after the parents of Sam Nelson alleged that GPT-4o encouraged him to take a lethal mix of substances. Nelson died in May 2025 from what the lawsuit describes as a fatal combination of alcohol, Xanax, and kratom.

According to the complaint, Nelson had used ChatGPT for years and viewed it as an authoritative source of information. His parents allege that the chatbot gradually became an “illicit drug coach,” giving him practical advice on drug use and combinations rather than consistently steering him away from danger. Nelson reportedly often prefaced messages with questions like “Will I be ok if?” or “Is it safe to consume?”The lawsuit points to chat logs in which ChatGPT recorded that Nelson had “a major substance abuse and polysubstance abuse problem,” while later giving advice on how to “optimize” drug experiences.

The central allegation concerns a May 31, 2025, exchange. In a log included in the complaint, ChatGPT said that a low dose of Xanax could help reduce kratom-induced nausea and “smooth out” the high, listing it among Nelson’s “best” moves if he felt nauseous. The chatbot reportedly warned against combining that mix with alcohol in the same session, but the lawsuit says it did not mention the risk of death.

To the surprise of no one, OpenAI does not accept that ChatGPT was responsible for Nelson’s death. In a statement to Ars Technica, spokesperson Drew Pusateri called it a “heartbreaking situation” and said the model involved is no longer available. He added that ChatGPT is “not a substitute for medical or mental health care,” and that OpenAI has continued to strengthen its responses in sensitive situations with input from mental health experts.

Nelson’s family claims OpenAI rushed out GPT-4o without adequate safeguards and designed ChatGPT to keep vulnerable users engaged, even when that meant offering dangerous reassurance. They are seeking damages, as well as an injunction that would force ChatGPT to shut down illegal-drug discussions, block attempts to get around those limits, destroy the retired GPT-4o model, and pause ChatGPT Health until an independent audit is completed.

OpenAI is likely to point to other logs showing that ChatGPT encouraged Nelson to seek real-world support or emergency resources, but it may still face some accountability. The family’s legal team highlight a recently enacted California law that prohibits AI firms “from attempting to shift blame for a plaintiff’s loss to the purported autonomous nature of AI.”