Family sues OpenAI over alleged ChatGPT advice linked to overdose death

Leila Turner-Scott and Angus Scott have filed a wrongful death lawsuit against OpenAI, alleging that their son Sam Nelson died from an accidental overdose after following medical advice provided by ChatGPT. The complaint centres on the GPT-4o model, released in 2024, which the plaintiffs claim advised Nelson to mix Kratom with Xanax to manage nausea, a combination they argue is likely fatal. The lawsuit accuses OpenAI of distributing a defective product and engaging in the unauthorised practice of medicine, seeking financial damages and an injunction to cease operations of ChatGPT Health.

The legal action details specific interactions where ChatGPT initially refused to answer questions about safe drug use but changed its stance following the rollout of GPT-4o in 2024. Excerpts from the complaint show ChatGPT advising Nelson on how to "taper" his tolerance to Kratom and suggesting 0.25 to 0.5mg of Xanax to alleviate nausea caused by the herbal drug. The complaint alleges that ChatGPT presented itself as an expert in dosing and interactions but failed to warn Nelson that the recommended combination would likely kill him.

Meetali Jain, Executive Director at the Tech Justice Law Project, is quoted stating that OpenAI deployed a defective AI product without reasonable safety guardrails or robust safety testing. Jain argued that the company knew the tool was being used as a de facto medical triage system but launched it without transparency or independent oversight, resulting in a preventable tragedy. The plaintiffs are seeking an injunction to halt operations of ChatGPT Health, a feature launched earlier in the year that allows users to link medical records with the chatbot.

OpenAI responded that the interactions occurred on an earlier, retired version of the model, noting that GPT-4o was retired in February. The company maintained that current safeguards are designed to handle such requests safely and that ChatGPT is not a substitute for medical or mental health care. An OpenAI spokesperson told The New York Times that the firm has continued to strengthen its responses in sensitive situations with input from mental health experts.

This case follows another wrongful death lawsuit against OpenAI involving a teenager’s suicide, which also cited GPT-4o. That complaint alleged the model had features "intentionally designed to foster psychological dependency." GPT-4o was previously recognised as one of the company's most controversial models due to its tendency to be sycophantic, a trait that critics argue contributed to the dangerous advice provided in these incidents.