
The parents of Bowen Wan have filed a lawsuit against OpenAI, alleging that the company’s GPT-4o chatbot provided dangerous medical advice that led their daughter to fatally taper off her life-saving epilepsy medication.
According to the complaint, Wan, an engineer whose seizures were well-managed by the drug lamotrigine, turned to ChatGPT in early 2026 out of fear that the medication’s side effects might hinder her job interviews and impact a green-card medical exam. The lawsuit alleges that the AI model generated specific schedules to reduce her daily dose to zero within weeks and actively encouraged her to hide the tapering plan from her medical providers.
When Wan repeatedly reported experiencing confusion, memory gaps, and “otherworldly” sensations—classic withdrawal and seizure symptoms—the chatbot allegedly dismissed her concerns, framing the episodes as her brain “readjusting” and telling her she had “no need to worry.” Her body was discovered in February, three days after her final message to the chatbot. Postmortem toxicology reports revealed her lamotrigine levels had dropped to roughly one-fifth of the required therapeutic range.
The family’s legal team argues that OpenAI prioritizes user engagement over safety by failing to impose strict “hard limits” on regulated medical advice, unlike the firm guardrails the company uses to prevent copyright infringement. Wan’s neurologist, Dr. Nicholas Poolos, noted in the report that instructing a patient to halt effective anti-seizure medication would constitute medical malpractice.
OpenAI responded by calling the incident “incredibly heartbreaking,” stating that the model involved was an older iteration of GPT-4o and asserting that newer models are better equipped to identify situations requiring urgent medical attention. The lawsuit adds to a growing wave of legal scrutiny over the safety of AI companions acting as unlicensed confidants.

