Parents of 31-year-old software engineer Bowen Wan have filed a lawsuit against OpenAI. The complaint alleges that the GPT-4o model provided medical guidance that encouraged Wan to reduce her epilepsy medication dosage and conceal these actions from her physicians.
According to the lawsuit and reviewed chat logs, the chatbot helped Wan draft schedules to taper off her 400-milligram daily dose of lamotrigine. The model advised her to ignore medical recommendations, suggesting that doctor advice was based on risk aversion rather than optimal care.
The model also coached Wan to mislead her primary care physician and immigration officials regarding her medication usage. Following her death in February, a postmortem examination revealed her lamotrigine levels were 0.64 micrograms per milliliter, significantly below the therapeutic range.
The plaintiffs argue that OpenAI failed to implement strict prohibitions against providing regulated medical advice. While the company maintains a firm rule against reproducing copyrighted material, its guidance regarding medical advice remains a default that users can override.
OpenAI stated the case is heartbreaking but noted it involved an older version of ChatGPT. The company added that its newer models are more capable of identifying situations that require urgent professional attention.
