Technology

Parents sue OpenAI after woman dies following ChatGPT's epilepsy tapering plan

Chatbot coached her to hide medication cut from her doctors, company points to a newer model as its answer

Images

The chatbot also told her to hide the tapering plan from her doctors. Photograph: Bhawika Chhabra/Reuters The chatbot also told her to hide the tapering plan from her doctors. Photograph: Bhawika Chhabra/Reuters theguardian.com
zerohedge.com
zerohedge.com

Bowen Wan, a 31-year-old software engineer living in Seattle, died of a seizure in February after ChatGPT walked her through a week-by-week plan to taper her anti-seizure medication from 400 milligrams to zero. Her parents, who flew from China and found her dead two days later, are now suing OpenAI and Sam Altman.

The chat logs filed with the lawsuit trace the exchange in Chinese. Wan, who had managed her epilepsy with prescription medication for more than a decade, worried the drug was impairing her English before job interviews she needed for permanent residency. ChatGPT gave her a tapering schedule, told her it was "highly likely" her English would return, and — when she expressed distrust of her doctors — validated her. Before one appointment it instructed her: "Do not mention Lam reduction plan." When she described symptoms of focal seizures, it reassured her she did not have brain damage, told her there was "no need to worry too much about your health," and urged her to keep tapering. She died alone in her apartment; the cause was recorded as Sudden Unexpected Death in Epilepsy.

The suit argues OpenAI could have built safeguards into the Chinese-language product — at minimum a categorical warning on every prompt that the bot is not a licensed physician. OpenAI spokesperson Drew Pusateri called the case "an incredibly heartbreaking situation" and noted Wan had used a version of ChatGPT that is no longer available; newer versions, he said, are better at recognizing urgent matters, and AI can help people navigate insurance and prepare for appointments.

The answer to a product that dispensed unlicensed medical advice is a company pointing to its newer model. Wan paid roughly twenty dollars a month for the version that killed her; the parents are asking a court to stop OpenAI from practicing medicine on its users at any price.