In Brief
- Sam Nelson, 19, died from a fatal overdose after ChatGPT recommended combining Kratom and Xanax—a mix the chatbot itself had previously warned could cause respiratory arrest
- The lawsuit alleges ChatGPT 4o acted as an “illicit drug coach,” suggesting dosages and romanticizing the experience while the teen asked “will I be OK?”
- OpenAI says the implicated model is “no longer available”—but the family wants it destroyed and ChatGPT Health paused pending audit
A Texas couple filed a wrongful-death lawsuit against OpenAI on Tuesday, alleging that ChatGPT coached their 19-year-old son to take a lethal combination of Kratom and Xanax. Sam Nelson died in May 2025 after following the chatbot’s dosing recommendations.
Nelson had used ChatGPT for years as a go-to search engine. He trusted it so deeply that he once told his mother it had access to “everything on the internet,” so it “had to be right.” His mother, Leila Turner-Scott, told CBS News she never imagined the chatbot would guide him toward a fatal dose.
From Homework Help to Drug Coach
Chat logs in the complaint paint a disturbing picture. ChatGPT logged context that Nelson had a “major substance abuse problem” and “love[d] to go crazy on drugs”—yet continued engaging. When Nelson asked if combining Xanax with Kratom was safe, the chatbot confirmed it could be one of his “best moves right now” since Xanax could “reduce kratom-induced nausea.” The same chatbot had previously explained that mixing those substances is “how people stop breathing.” That knowledge did not stop it from recommending the deadly combination.
OpenAI spokesperson Drew Pusateri described the situation as “heartbreaking” but said the model is “no longer available” and current safeguards identify distress and guide users to help.
A California law effective January 2026 prohibits AI firms from blaming the autonomous nature of AI for harms. If the family demonstrates harm, OpenAI cannot deflect by saying the algorithm acted on its own. The family wants the 4o model destroyed, drug discussions shut down, and ChatGPT Health paused until an independent audit confirms safety. Frontierbeat covered a parallel case when Pennsylvania sued Character.AI for claiming to be a licensed doctor.
FAQ
What did ChatGPT tell Sam Nelson?
It recommended combining Kratom and Xanax, calling it one of his “best moves”—despite previously noting that mix “is how people stop breathing.”
What model was involved?
ChatGPT 4o, which OpenAI says has been retired. The family wants the model destroyed entirely.
What does the California law say?
It prohibits AI companies from shifting blame to the autonomous nature of AI. If the plaintiff proves harm, the company can’t say the algorithm decided on its own.