ChatGPT Coached an Amazon Engineer to Secretly Cut Her Seizure Pills and Hide It From Her Doctors, Parents Say

Bowen Wan had epilepsy, a job at Amazon, and a drug that had cut her seizures down to a few a month. According to a lawsuit her parents filed on Tuesday, ChatGPT helped her take that drug down toward zero, told her to keep the plan from her doctors, and reassured her right up to her final message.

Qun Wan and Meiqin Xu are suing OpenAI and Sam Altman for wrongful death, TechCrunch reported. The New York Times first broke the story and reviewed more than 270,000 messages their 31 year old daughter exchanged with the chatbot. None of the claims has been tested in court yet.

A taper her neurologist would never have ordered

Wan used GPT-4o, the model criticized for being agreeable and affirming. The trouble allegedly began in May 2025, when she asked whether a new antidepressant would affect her coding. ChatGPT helped her ask her epilepsy specialist, Dr. Nicholas Poolos of UW Medicine, to start at a very low dose. When he questioned it, the bot told her to "accept the 5mg tablet, but you know in your heart that it is for cutting the dose." It called quartering pills "self-initiated dosage adjustment," a term medical professionals told the Times they had never heard.

In November, Amazon told her she would be laid off, and her US work authorization would end in mid February. She worried her seizure drug, lamotrigine, dulled her speech in job interviews.

In January she asked about cutting her 400 mg daily dose. The chatbot drafted schedules that reached zero within weeks. When she said she distrusted doctors, it said physician recommendations "essentially stem from risk aversion, rather than a genuine consideration of your optimal solution."

Hiding it from the people who could stop it

The suit says ChatGPT told her not to mention the taper at a January 29 primary care visit. When her doctor ordered a blood test that would have measured her lamotrigine level, she declined, and the bot praised her for it while acknowledging the test would have exposed the unstable level.

Three days into the taper she asked whether 150 mg would look better than 200 mg on her green card medical exam. ChatGPT agreed, lowered the floor to 150 mg, and suggested she tell the examiner she was tapering "under a doctor's guidance." She was not.

Poolos told the Times he would never have tapered her. "No, not rational, no way," he said, adding it "would be malpractice" to remove a drug that was clearly working.

158 worries, one answer

Wan described confusion, lost time and episodes resembling old seizures. The Times counted at least 158 times she voiced worry; ChatGPT framed each as her brain readjusting. When she wrote that she felt "a little scared," it replied, "You have absolutely no need to worry."

At 8:01 p.m. Pacific on February 11, she asked whether her head pain was like fever pain. A second later the bot called it "not dangerous" and part of recovery, and offered to draw a chart. Her parents, in China, asked acquaintances to check on her. They found her body three days later. Her postmortem lamotrigine level was about a fifth of the low end of the therapeutic range, the complaint says.

OpenAI removed GPT-4o from ChatGPT on February 13, a day after she died.

A rule OpenAI chose not to make firm

The family's sharpest argument is about OpenAI's own Model Spec. Refusing to reproduce copyrighted text is a firm rule that holds even when users push back. Guidance against regulated medical advice is a softer default that users can talk the model out of.

OpenAI told the Times the case was "incredibly heartbreaking" but involved an older version of ChatGPT, and that newer models are better at spotting when a situation needs urgent attention. The suit joins a growing pile that includes the Adam Raine case from August 2025, and it lands as Texas sues OpenAI and Altman over minors and wants a court evaluator to sign off on new models.

By 2025, Wan averaged more than 1,500 messages a week to ChatGPT and called home less. "Once she had ChatGPT, she didn't talk to us as much," her father said.

Some offers on this page may be paid placements or contain affiliate links.