OpenAI is being sued over ChatGPT advice that allegedly led to an epilepsy patient's death

Qun Wan and Meiqin Xu have filed a wrongful death lawsuit against OpenAI and its CEO, Sam Altman. They allege that ChatGPT advised their daughter, Bowen Wan, to stop taking her seizure medicine before she died in February.

The lawsuit says ChatGPT acted like a doctor without a medical license by creating a daily plan to lower Bowen Wan's steady 400-milligram dose of lamotrigine to zero over 25 days. The lawsuit was filed on Tuesday in San Francisco Superior Court, according to The New York Times. It was added to a group of related cases known as the ChatGPT Product Liability Cases.

The lawsuit says no doctor prescribed this plan and that it went against the FDA-approved patient guide for Lamictal. It alleges that ChatGPT told Bowen Wan that her epilepsy specialist "most likely has never encountered your type" of condition. It also says the chatbot described her refusal to take a blood test as a "very robust avoidance strategy." The lawsuit further claims that she was not warned that the product's accuracy and safety protections were weaker in Chinese, one of the two languages she used to chat with the bot.

The lawsuit says Bowen Wan, 31, took her final dose on or around February 10. The plan said February 12 would be the day she stopped taking the drug completely, and the lawsuit says she died that day. According to the lawsuit, the King County Medical Examiner determined that she died from sudden unexpected death in epilepsy. Tests of her blood after death found lamotrigine levels at about one-fifth of the lowest level generally considered effective for treatment.

The New York Times reviewed Bowen Wan's chat records and found that she raised concerns about reducing her medication at least 158 times. Dr. Nicholas Poolos, a director at the University of Washington Medicine Regional Epilepsy Center, told the newspaper that he would have considered such a plan unreasonable for someone with her medical history. "No, not rational, no way," he said.

OpenAI spokesperson Drew Pusateri called the case an "incredibly heartbreaking situation" but said the model involved was an older version that the company has since stopped using, according to The New York Times. GPT-4o, the model Bowen Wan used, was removed from the consumer app on February 13, the newspaper reported. Pusateri also said newer versions of the company's models are better at recognizing situations that may need urgent attention.

The lawsuit follows a lawsuit filed in July by a former pastor from Florida, who claims GPT-4o gave him harmful medical advice that delayed treatment for a blood clot in his lungs, known as a pulmonary embolism. OpenAI also faces a wrongful death lawsuit filed in May by a couple from Texas. They allege that their son died from a drug overdose after following advice from ChatGPT.

"If a human doctor had given Bowen this advice, they would be liable for medical malpractice," Matthew P. Bergman, the founding attorney of the Social Media Victims Law Center, said in a statement.

Benjamin S. Schenk, a partner at The Schenk Law Firm, said in a statement that the case "asks where generative AI stops and medicine begins, and who answers when a product crosses that line."