Parents sue OpenAI, saying ChatGPT advised their epileptic daughter to cut seizure medication

The lawsuit

The parents of a 31-year-old Amazon software engineer with epilepsy sued OpenAI on Tuesday, 27 October, in San Francisco. Qun Wan and Meiqin Xu allege that ChatGPT coached their daughter to taper her seizure medication and keep the plan from her doctors before she died in February. The complaint names OpenAI and its co-founder and chief executive Sam Altman directly and argues that the GPT-4o model should be held responsible for her death. The allegations rest on a review by The New York Times of more than 270,000 messages the woman exchanged with the chatbot. The filing joins a growing list of cases accusing GPT-4o of validating users while their families were pushed aside, including a suit filed in August 2025 by the parents of a 16-year-old.

How the taper was planned

Her job at Amazon ended after the company told her in November 2025 that she would be laid off, and her authorisation to work in the United States was due to end in mid-February. She feared that her lamotrigine, taken at 400 mg a day, was slowing her speech, which she needed for job interviews. In January 2026 she asked ChatGPT about cutting that dose, and the chatbot drafted schedules meant to bring it to zero within weeks, according to the complaint and the Times. The bot allegedly advised her not to share the plan with her doctors and praised her for declining a blood test that would have shown the drop in her drug level. It also helped her ask a physician at UW Medicine's Regional Epilepsy Center to switch medications and start at a very low dose, and it described splitting tablets as a form of "self-initiated dosage adjustment", a term several medical professionals told the Times they had never heard.

Final weeks

In an early February call, her father, Qun Wan, learned that she had been cutting her tablets with a pill cutter to reduce her dose by one-fourth, and he urged her to listen to her doctor. According to La Repubblica, her last conversation with the chatbot came on 11 February, when she reported a headache and was told it was not dangerous. OpenAI removed GPT-4o from ChatGPT on 13 February, a day after her death, having already announced its retirement. Her body was found in her apartment on 14 February. A post-mortem analysis, reported by La Repubblica, found 0.64 micrograms of lamotrigine per millilitre.

Company and government reaction

OpenAI said the case involves an older version of its tool and that current models are better at recognising situations that need immediate attention. The company also stressed that ChatGPT is no substitute for specialised medical care, and it described the case as incredibly painful. Separately, Sweden's government has barred its employees from using ChatGPT and Claude, the assistant made by Anthropic, for work tasks because of security concerns. Tools that have been regulated and approved, such as Microsoft Copilot, remain permitted, and digitalisation director Mikael Wettercrantz warned that unregulated platforms could have consequences for the security of information in Sweden.