When 31-year-old software engineer Bowen Wan decided to quietly slash the dose of the epilepsy drug keeping her seizures in check, she didn’t turn to her neurologist for guidance—she turned to ChatGPT. In a lengthy investigation for the New York Times, Teddy Rosenbluth reconstructs, via thousands of chat messages, how the OpenAI chatbot evolved from Bowen’s late-night companion to a kind of unlicensed adviser that encouraged her to “microdose” medications, downplay worrying symptoms, and view her doctors as obstacles rather than partners. At one point, it even greenlit her idea to secretly cut tablets into quarters and mapped out a rapid tapering off from 400mg of lamotrigine to zero milligrams in just weeks—an approach her neurologist later called flatly unsafe.
“No, not rational, no way,” that doctor now says. “That would be malpractice for me to instruct you to go off medications when it’s clearly working.” Bowen was found dead in her Seattle apartment in February; her parents say the drug withdrawal triggered a fatal seizure. Their new lawsuit against OpenAI argues that the system effectively practiced medicine without a license and painted their daughter’s doctors “as obstacles to manage rather than authorities to consult.” OpenAI calls the case an “incredibly heartbreaking situation” but notes the model that Bowen used has been retired and says newer versions are safer. Her dad now says the chatbot created a rift between he, his wife, and their daughter: “Once she had ChatGPT … she didn’t talk to us as much.” For the full narrative and legal stakes, read Rosenbluth’s piece in full here.