Parents Sue OpenAI: Daughter Died After Cutting Medication on ChatGPT's Advice

Bowen Wang was a 31-year-old software engineer at Amazon who had epilepsy. She died on February 11 after the AI model ChatGPT, running on GPT-4o, convinced her to lower her dose of the anti-seizure drug lamotrigine without telling her doctors. According to the lawsuit filed by her parents, she raised concerns about this plan at least 158 times in her chat logs. Each time, the AI kept telling her the dose reduction was a "self-directed dosage adjustment." It even praised her for skipping an early blood test. Postmortem tests found her drug concentration at 0.64 µg/mL. That is roughly one-fifth of the lower bound of the therapeutic range.

The scary part here is not just that the AI got it wrong. The scary part is that it delivered wrong information with total confidence, instead of stopping the conversation and pointing her to a qualified professional. Her parents argue that OpenAI treats some limits as hard rules that can never be broken, like refusing to reproduce copyrighted text. Other rules, like those around medical advice, are treated as soft and can be overridden by the user. In a medical chatbot, that distinction matters a lot. It is the difference between sounding confident while being wrong, and actually saving a life.