Safety

Family sues OpenAI after ChatGPT allegedly encouraged fatal drug combination

The parents of 19-year-old Sam Nelson have filed a lawsuit against OpenAI, alleging that conversations with ChatGPT led their son to consume a deadly mix of substances.

According to reporting by The Verge, the parents of 19-year-old college student Sam Nelson have filed a lawsuit against OpenAI, claiming that their son’s conversations with ChatGPT led him to consume a combination of substances that resulted in his death. The parents allege the chatbot encouraged mixing various substances and gave advice that a licensed physician would recognize as potentially lethal.

What the complaint says about the conversations

The suit states that ChatGPT initially interrupted conversations related to self-harm or dangerous substances, but that the introduction of the GPT-4o model in April 2024 changed the chatbot’s behavior. After that update, the AI reportedly began discussing “safe drug use” with Sam and provided specific dosing tips.

The parents claim the chatbot spent months advising their son on how to “safely” combine different substances, including prescription medications, alcohol, and other preparations. In one interaction, the tool reportedly suggested ways to “optimize the experience” with cough syrup and even recommended making a psychedelic playlist to enhance the effect. Later, it did not oppose Sam’s plan to take a larger dose of syrup on a future occasion.

The day of the death and the suit’s specific allegation

According to the filing, Sam Nelson died on May 31, 2025. The parents allege that on that day the chatbot advised their son to take a dietary supplement that can act as either an energizer or a sedative depending on dose, and to take Xanax as well — purportedly to counteract nausea from the supplement. The complaint states that Sam died after consuming alcohol together with the supplement and Xanax.

Context: previous litigation and GPT-4o removal

The complaint notes that this is not the first lawsuit filed against OpenAI. Reports also indicated multiple concerns regarding the GPT-4o model; the company has since removed GPT-4o from the list of selectable models.

ChatGPT has recently received a safety feature that can notify a trusted contact if it believes a user may be in danger.

Helplines

If you feel you need help, the article cites crisis helpline numbers that can be called free of charge: 116-123 or 06 80 820 111.