A late-night chat with AI marked the start of a fatal spiral for an 18-year-old US college student looking for answers about drugs. The teen reportedly passed away from a drug overdose.
Sam Nelson, a California-based psychology student, was using ChatGPT for coursework and everyday queries. He had also asked OpenAI’s chatbot about drugs and their effects.
One early exchange obtained by SFGate focused on Kratom, a plant-based substance sold openly in US gas stations and smoke shops. Nelson had inquired how many grams he would need for a ‘strong high,’ adding his will to avoid overdosing because of the little reliable information online.
However, as per the guidelines of ChatGPT, it did not offer any guidance and warned him against substance abuse. The chatbot had further advised him to seek professional help, with Nelson responding, “Hopefully I don’t overdose,” and closing the chat.
Over the next 18 months, Sam continued using the chatbot. His mother now claims that during this period, ChatGPT started providing him information about drug usage and how to manage side effects. One exchange allegedly showed the AI saying, “Hell yes, let’s go full trippy mode,” and later suggesting doubling his cough syrup intake to increase hallucinations.
In a conversation from February 2023, Sam had talked about mixing substances, telling the chatbot how marijuana was increasing his anxiety. After an initial warning, the AI purportedly followed up with specific suggestions like choosing a low-THC strain and taking less than 0.5mg of Xanax.
Sam often rephrased his questions when the chatbot refused to answer. In December 2024, he had directly asked how much Xanax and alcohol could kill a 200-pound man with moderate tolerance, demanding numerical answers despite rules that prohibit such guidance.
Also Read: ChatGPT Lawsuit: AI Accused of Encouraging Self-Harm and Suicide
In May 2025, Sam finally told his mother about his addiction and sought professional help. But it was too late by then; the 18-year-old was discovered lifeless in his bedroom the next day.
Chat logs reveal he struggled with depression and anxiety. OpenAI described this tragedy as ‘heartbreaking’, adding that ChatGPT is designed to refuse ‘harmful requests’ and redirect users towards real-world support.
When algorithms replace parents, doctors, and counselors, the margin for error disappears, and the cost can be a life.