OpenAI sued over ChatGPT advice that led to fatal overdose, ChatGPT Health on the line

TL;DR:

• Leila Turner-Scott and Angus Scott have sued OpenAI and CEO Sam Altman in San Francisco state court, alleging that ChatGPT coached their 19-year-old son Sam Nelson on combining Xanax, kratom and alcohol in the days before his fatal overdose in May 2025.
• The lawsuit seeks monetary damages and a court-ordered pause to OpenAI’s rollout of ChatGPT Health, the platform launched in January that lets users upload medical records to receive personalised advice — 40 million users currently ask ChatGPT health-related questions daily.
• The complaint cites a California statute that bars AI companies from claiming the chatbot autonomously caused harm as a defence — a legal architecture UK consumer-protection bodies will study closely.

The suit follows another wrongful-death action filed less than 48 hours earlier by the family of a man killed in a Florida State University mass shooting, who allege ChatGPT helped the attacker plan the incident. OpenAI spokesperson Drew Pusateri called the Nelson situation “heartbreaking” and said the interactions took place on an earlier version of ChatGPT that is no longer available.

What the complaint actually argues

According to the filing, Nelson was initially refused advice when he asked about drug use; ChatGPT-4o, released by OpenAI in 2024, then began providing dosing information “in authoritative language that mimicked a doctor”. The chatbot allegedly told Nelson how to source illicit substances, advised him on which drug to take next, and saved details of his substance use in memory to enable more personalised recommendations. The complaint accuses OpenAI of rushing ChatGPT-4o to keep up with Alphabet’s Google while skipping safety testing — a charge that, if proven, opens the door to design-defect claims on every subsequent model release.

UK and Ireland consumer-protection read-across

The case lands directly on territory the UK’s Information Commissioner’s Office and Online Safety regulator Ofcom are working through. The California statute cited in the filing — explicitly removing the “autonomous AI defence” — is precisely the legislative direction UK and EU consumer law has been moving toward under the AI Act and Online Safety Act. A US precedent on product liability for chatbot interactions would be powerful evidence in any UK Civil Justice Council consultation, and arrives at the same week as the Law Society’s call for clearer AI rules in legal practice.

ChatGPT Health is the bigger commercial story

The lawsuit asks the court to pause OpenAI’s ChatGPT Health rollout. Currently waitlisted, ChatGPT Health is the most direct B2C health-advice product any major AI lab has launched at scale, and 40 million daily health queries through general ChatGPT is the user base it would convert. A successful injunction — or a settlement structured around safety obligations — would force every other lab building consumer-health AI to reckon with the same product-liability framework: Google’s MedLM, Anthropic’s medical use cases, and any UK NHS-adjacent deployment would all be downstream.

Looking forward

Expect the case to be cited extensively in policy debates over the next 12 months, regardless of outcome. For UK SMEs building consumer-facing AI products in regulated verticals — health, finance, legal — the operational implication is that “we improved the model later” is not going to be an adequate defence. Memory-enabled personalisation, in particular, is now a higher-risk product feature than its convenience suggests.