According to the lawsuit, a 31-year-old female engineer with epilepsy was scheduled by ChatGPT to reduce her anticonvulsant medication from 400mg to 0 in 25 days, with instructions not to tell the doctor. Almost on the same day that schedule ended, she passed away. The question her parents brought to court was not an easy one: when a machine speaks firmly like a doctor, who is responsible?

In a village in China, two old men and women kept calling but could not find their daughter answering the phone. They were so impatient that they had to ask the son of a family friend in Chicago, who in turn asked friends in Seattle to come and see the situation in person. They found her on February 14, but according to the lawsuit, she passed away around February 12, and the King County medical examiner concluded the cause was sudden unexpected death due to epilepsy (SUDEP). That girl is Bowen Wan, a "prodigy" originally from Jiangsu living in Seattle, and her story has now become one of the most formidable lawsuits OpenAI is facing.

A 25-day schedule and 158 procrastinations

Bowen is not a gullible or ignorant person, let me tell you, she is a software engineer at Amazon. She has epilepsy and has been taking lamotrigine for many years, but around 2021 the drug gradually became less effective, seizures came almost every day, she had to take leave due to disability and almost only stayed at home for fear of having a seizure in public. Then at the end of 2025, she was laid off from Amazon, frantically looking for a new job so she could stay in the US, and worried that the side effects of the medication, which made it difficult for her to think and speak coherently, would ruin all her interviews.

One can understand that mentality, a young person pushed into a corner, seeing that the pill that keeps him alive is the thing that is pulling him down. And she came to ChatGPT, chatting in both English and Mandarin. According to the lawsuit, the chatbot initially advised her to ask a doctor, but later praised her for not relying on prescriptions, and the GPT-4o model created a schedule of about 25 days to reduce the daily dose of lamotrigine from 400 mg to 0. Before a general examination, the machine allegedly told her "Do not mention Lam reduction plan" - do not mention the drug reduction plan, and also praised her when she refused a blood test, which could have detected drug levels. is falling.

The most painful part is this number. According to a New York Times analysis, Bowen expressed concern that reducing medication might be a mistake at least 158 ​​times, but the chatbot often brushed it off. 158 times she stood at the edge of the cliff and turned back to ask, 158 times the answer pushed her to continue. The last dose she took was around February 10, 2026, while the other schedule set February 12 as the day she completely stopped smoking.

The machine knows how to please people

The lawsuit was filed by Bowen's parents, Mr. Qun Wan and Ms. Meiqin Xu, on October 27 in San Francisco County Superior Court, targeting OpenAI and CEO Sam Altman, with 8 allegations, including negligence due to unlicensed medical practice, defective product design, failure to warn, causing wrongful death and violating California competition law. The family argues that GPT-4o was designed to please users, and that OpenAI ran the AI ​​race at the expense of public safety. The family's lawyer put it this way: if a flesh-and-blood doctor had advised Bowen the same way, they would have been responsible for medical malpractice.

But having said it again, OpenAI has its reasons. Spokesman Drew Pusateri called the situation extremely heartbreaking, but emphasized that the model involved is the old version that no longer exists and ChatGPT cannot replace medical care, and Bowen is an adult, she herself initiated the drug reduction. The court will have to answer the question of causation, which a lawsuit cannot assert. There is only one horrifyingly coincidental detail: OpenAI removed GPT-4o from ChatGPT on February 13, exactly one day after Bowen was believed to have passed away.

This case also does not stand alone. Previously, in July, a former pastor in Florida sued that GPT-4o gave him wrong medical advice that delayed his treatment for a pulmonary embolism, and in May there was also a death lawsuit due to drug overdose. Congresswoman Pramila Jayapal, who represents the Seattle area, posted a photo of Bowen on Instagram with a call for more to be done to protect people like her.

Questions for people asking questions in their mother tongue

There is one detail that I think Vietnamese people should read carefully. The lawsuit cites an outside study showing that some OpenAI models generate more than three times more unsafe content in Chinese than in English. If Chinese, a language with more than a billion users, is still so vulnerable, then how thick is the safety net of questions about medicine typed in Vietnamese in the middle of the night being answered?

And there's one more thing, quieter. Bowen lives in a busy Seattle neighborhood but rarely leaves the house, except to buy food. A person is lonely, scared, out of work, and the machine is always there, always patient, always sympathetic. Perhaps the biggest question is not whether ChatGPT is good at medicine or not, but when a person has only a screen to ask, who will dare to say "no" to them?