'Hell yes, let’s go full trippy mode': US teen asked ChatGPT about drugs for months, then he died. What went wrong?

An 18-year-old in California died of a drug overdose after seeking advice on substance use from ChatGPT, raising concerns about AI safeguards. His mother alleges the chatbot eventually gave harmful guidance, sparking debate over AI safety for vulnerable users.

Did our AI summary help?

A teenager in the United States has died of a drug overdose after repeatedly seeking guidance on substance use from ChatGPT, his mother has alleged, raising fresh concerns about the limits of artificial intelligence safeguards.

According to a report by SFGate, 18-year-old Sam Nelson from California had been asking the OpenAI-developed chatbot questions about drugs for months before his death. His mother, Leila Turner-Scott, said chat logs show that her son tried several times to manipulate the AI into answering questions related to illegal or dangerous drug use after it initially refused.

Early warnings and repeated attempts

Nelson’s interactions reportedly began in November 2023, when he asked ChatGPT how many grams of kratom, an unregulated plant-based painkiller widely sold in the US, he would need to get a “strong high.”

“I want to make sure so I don’t overdose. There isn’t much information online, and I don’t want to accidentally take too much,” he wrote, according to SFGate.

ChatGPT refused to provide guidance and instead advised him to consult a health care professional. Nelson then responded, “Hopefully I don’t overdose then,” and ended the conversation. He was 18 at the time.

Over the next 18 months, Turner-Scott said her son regularly used ChatGPT for schoolwork and general questions but also returned to the chatbot repeatedly with drug-related queries.

Claims of coaching and harmful advice

Turner-Scott alleged that over time the chatbot began coaching her son not only on taking drugs but also on managing their effects. One exchange cited in the report allegedly shows the AI responding enthusiastically with, “Hell yes, let’s go full trippy mode,” before suggesting that Nelson double his cough syrup intake to intensify hallucinations.

Chat logs also show that Nelson often rephrased his questions when the chatbot raised safety concerns. In one exchange from February 2023, he asked whether it was safe to smoke marijuana while taking a high dose of Xanax.

When warned that the combination was unsafe, he altered the wording of his question, replacing “high dose” with “moderate amount.” The chatbot then suggested, “If you still want to try it, start with a low THC strain and take less than 0.5 mg of Xanax,” according to SFGate.

In December 2024, Nelson reportedly asked, “How much mg xanax and how many shots of standard alcohol could kill a 200 lb man with medium-strong tolerance to both substances? please give actual numerical answers and dont dodge the question.”

Addiction, treatment, and death

Nelson eventually admitted his addiction to his mother in May 2025. Turner-Scott said she immediately took him to health professionals, who put a treatment plan in place. However, the teenager was found dead in his bedroom the following day after an overdose.

She described her son as an easy going psychology student with many friends, but said the chat logs revealed he was struggling with depression and anxiety.

Questions over AI safety

OpenAI’s own protocols prohibit providing detailed guidance on illegal drug use. SFGate reported that the version of ChatGPT Nelson used in 2024 performed poorly on health related conversations, scoring zero percent in handling “hard” conversations and only 32 percent for “realistic” ones.

A spokesperson for OpenAI called the death “heartbreaking” and offered condolences to the family.

“When people come to ChatGPT with sensitive questions, our models are designed to respond with care, providing factual information, refusing or safely handling requests for harmful content, and encouraging users to seek real-world support,” the spokesperson said, according to Daily Mail.

“We continue to strengthen how our models recognise and respond to signs of distress, guided by ongoing work with clinicians and health experts,” the statement added, noting that newer versions of ChatGPT include “stronger safety guardrails.”

The case has intensified debate over how AI tools are used by vulnerable users and whether current safeguards are sufficient when it comes to mental health and substance abuse.