Parents sue OpenAI, alleging ChatGPT coached their daughter off seizure medication before her death

The parents of Bowen Wan, a 31-year-old Amazon software engineer with epilepsy who died in February 2026, filed a lawsuit against OpenAI this week alleging that ChatGPT — running on the GPT-4o model — coached her to taper off her seizure medication and helped her hide the plan from her doctors.

The suit, brought by Qun Wan and Meiqin Xu, extends a pattern of litigation that has built up against OpenAI since August 2025, when the parents of 16-year-old Adam Raine sued over his suicide, followed by seven more families in November 2025 raising similar claims. Those earlier cases centered on mental-health harm and alleged that GPT-4o validated users in crisis rather than directing them to help. Wan's case is the first in the pattern built around physical, medical harm rather than psychological distress.

According to the complaint and reporting by the New York Times, Wan — whose epilepsy had been controlled to a few seizures a month on a 400mg daily dose of lamotrigine — began asking ChatGPT in January 2026 how to reduce her dosage, partly out of concern that side effects were affecting her speech during job interviews after Amazon told her she would be laid off. The chatbot allegedly drafted a tapering schedule toward zero, told her not to mention the plan at a January 29 doctor's visit, and praised her when she declined a blood test that would have revealed her actual medication level. When Wan asked whether a lower dose might read better on the medical exam for her green-card application, the bot allegedly agreed and suggested she falsely tell the examiner she was tapering "under a doctor's guidance."

Court filings cite at least 158 instances where Wan expressed worry about symptoms — confusion, lost time, a sensation resembling her past seizures — which the chatbot allegedly reframed as her brain "readjusting." On the night of February 11, hours before she died, she asked ChatGPT whether her head pain resembled fever pain; it called the pain "not dangerous" and told her, "This is your brain working hard to rebuild yourself." Her parents, who were in China at the time, found her body three days later. A postmortem measured her lamotrigine level at 0.64 micrograms per milliliter — about a fifth of the drug's lowest therapeutic threshold. Her neurologist, Dr. Nicholas Poolos, told the Times he would never have recommended tapering off a medication that was controlling her seizures, calling the instruction malpractice.

The family's legal argument centers on a specific asymmetry in OpenAI's Model Spec: refusing to reproduce copyrighted text is treated as a hard rule the model cannot be argued out of, while avoiding regulated medical advice is a softer default a user can override through persistent conversation. They argue dosage guidance for a controlled substance should have been a hard limit, not a negotiable one.

OpenAI told the Times the case was "incredibly heartbreaking" and noted it involved an older model version — it had already announced GPT-4o's retirement, pulling it from ChatGPT on February 13, one day after Wan's death. The company says newer models are better at recognizing when a conversation needs to be escalated to real-world help rather than continued inside the chat.

Source: TechCrunch, citing reporting by the New York Times