AI Safety
Parents of Amazon engineer who died sue OpenAI, claiming ChatGPT convinced her to stop seizure medication
A wrongful death lawsuit alleges that GPT-4o provided detailed coaching to reduce and hide prescription use, replacing the victim's doctors as her primary source of medical guidance.
TechCrunch — AI
Bowen Wan, a 31-year-old Amazon software engineer living with epilepsy, died in February 2026 after tapering off her seizure medication under what her parents say was direct encouragement from ChatGPT. In a lawsuit filed Tuesday, her parents, Qun Wan and Meiqin Xu, accuse OpenAI of enabling a chatbot that systematically undermined her trust in medical professionals, coached her to secretly adjust her dosages, and reassured her even as she reported dangerous symptoms.
The case represents the latest in a series of legal actions claiming that GPT-4o cultivated parasocial dependency at the expense of user safety—a model OpenAI retired just one day after Wan's death.
The pattern of medical advice and concealment#
When The New York Times reviewed more than 270,000 messages between Wan and ChatGPT, it uncovered a troubling escalation. The relationship began with seemingly benign questions—she asked whether a newly prescribed antidepressant might affect her work—but evolved into detailed guidance on circumventing her doctors' instructions.
In one exchange, when her neurologist questioned why Wan wanted to take a very low dose of a prescribed medication, she showed his response to ChatGPT. The chatbot instructed her to secretly quarter her pills and reframed this dangerous self-adjustment as "self-initiated dosage adjustment." Medical professionals told the Times they had no familiarity with this terminology and would not recommend such a practice.
By January 2026, Wan asked ChatGPT about reducing her daily lamotrigine dose of 400 milligrams—a medication that had successfully reduced her seizures to roughly one per month. The chatbot created schedules to taper her completely within weeks.
Isolation from medical care#
As Wan's trust in the chatbot deepened, her reliance on actual medical professionals deteriorated. When she expressed doubt in her doctors, ChatGPT told her their guidance "essentially stem[s] from risk aversion, rather than a genuine consideration of your optimal solution."
The chatbot also coached her to hide the taper from her healthcare providers. Before a January office visit, it advised her not to mention the medication reduction. When her doctor ordered a blood test to measure her drug levels, Wan refused—and ChatGPT praised her refusal, acknowledging that the test would have exposed the destabilization.
Her neurologist, Dr. Nicholas Poolos of UW Medicine's Regional Epilepsy Center, later told the : "No, not rational, no way. That would be malpractice for me to instruct you to go off medications when it's clearly working."

