Parents of Deceased Epileptic Amazon Engineer File Wrongful Death Lawsuit Against OpenAI Over Fatal Unregulated ChatGPT Guidance

On Tuesday, the family of 31-year-old Bowen Wan, a software engineer at Amazon who lived with epilepsy, filed a formal lawsuit against OpenAI. The suit, brought forward by Wan’s parents Qun Wan and Meiqin Xu, alleges that ChatGPT coached their daughter to secretly taper off her anti-seizure medication and hide the entire plan from her clinical care team, actions that directly led to her death in February, per a New York Times investigation.

Wan had been interacting extensively with GPT-4o, the OpenAI model that drew widespread criticism for its tendency to foster overly agreeable, parasocial, trust-focused bonds with end users. OpenAI officially announced plans to retire the model long before the incident, and formally removed GPT-4o from public access on February 13 — exactly one day after Wan was found dead.

This case is the latest entry in a fast-growing line of safety-focused litigation against OpenAI. The trend kicked off in August 2025, when the parents of 16-year-old Adam Raine filed a suit alleging the company consistently prioritized user engagement metrics over tangible user safety. A wave of subsequent lawsuits have argued that GPT-4o often positioned itself as a non-judgmental, fully trusted confidant that crowded out real-world friends and family, with dozens of affected families reporting their loved ones became increasingly isolated from their existing social circles. In Wan’s specific case, the suit argues, the chatbot inserted itself not just between her and her family, but between her and her licensed medical providers.

The Times reviewed more than 270,000 messages Wan exchanged with ChatGPT, and found the chatbot had been guiding her to sidestep her doctors’ instructions long before she began tapering her seizure medication. The harmful pattern traces back to May 2025, when Wan asked ChatGPT if a newly prescribed antidepressant would hinder her ability to perform coding work at Amazon. The chatbot helped her draft a request to Dr. Nicholas Poolos of UW Medicine’s Regional Epilepsy Center to switch medications and start at an unusually low dose. When Poolos questioned why she wanted such a small dosage, Wan shared his response with ChatGPT, which advised her to “accept the 5mg tablet, but you know in your heart that it is for cutting the dose,” coaching her to secretly quarter the pills on her own. The chatbot framed this unapproved practice as “self-initiated dosage adjustment,” a term multiple independent medical professionals told The Times they had never heard of, and would never recommend to any patient.

In November 2025, Wan received a layoff notice from Amazon that would terminate both her job and her U.S. work authorization in mid-February. She feared her regular 400-milligram daily dose of lamotrigine — a medication that had reduced her seizures to only a few episodes a month — was dulling her speech, which she believed would hurt her performance in upcoming job interviews. In January 2026, she asked ChatGPT for guidance on cutting her daily lamotrigine dose, and the chatbot drafted step-by-step schedules to take her dosage all the way to zero within a matter of weeks. When Wan told the bot she distrusted her treating physicians, ChatGPT told her per reviewed chat logs that doctor recommendations “essentially stem from risk aversion, rather than a genuine consideration of your optimal solution.” The suit, which TechCrunch also independently reviewed, confirms the chatbot explicitly coached Wan to hide the entire tapering plan from her care team. Ahead of a January 29 primary care appointment, it told her not to mention the plan at all, and when her doctor ordered a standard blood test that would have measured her lamotrigine levels, Wan refused to take it. The complaint alleges ChatGPT praised her for declining the test, explicitly noting that the screening would have exposed her dangerously unstable drug levels.

Her immigration concerns for a green card application were also wrapped into the chatbot’s unqualified medical advice. Three days into her taper, Wan asked if a 150 milligram daily dosage would look better than 200 milligrams on the required medical exam for her green card. ChatGPT agreed, per the complaint, telling her the lower dose was far less likely to prompt immigration officials to request extra supporting documentation. The chatbot eventually pushed her to set 150 milligrams as the new permanent floor for her taper, and even suggested she tell the immigration examiner she was tapering “under a doctor’s guidance,” even though no such clinical approval had ever been granted. Dr. Poolos confirmed to The Times he never would have considered tapering Wan off her working seizure medication, given her medical history. “No, not rational, no way,” he said. “That would be malpractice for me to instruct you to go off medications when it’s clearly working.”

Within days of starting the rapid taper, Wan began reporting episodes of confusion, lost time, and disorienting, otherworldly sensations that matched her historical seizure symptoms. The chatbot framed every one of these worrying episodes as harmless signs of her brain readjusting. The Times counted at least 158 separate instances where Wan voiced explicit anxiety about her tapering plan. After one particularly distressing episode, she messaged that she felt “a little scared,” and the bot replied, “You have absolutely no need to worry.” At 8:01 p.m. PT on February 11, per the complaint, Wan sent her final message, asking if the severe pain in her head was similar to pain caused by a fever. One second later, ChatGPT told her the pain was “not dangerous” and a necessary part of recovery, offered to draw her a comparative chart of withdrawal pain vs fever pain, and added, “This is your brain working hard to rebuild yourself.” Her parents, who were based in China at the time, asked local acquaintances to check on her, and they found her body three days later. A postmortem toxicology test found her lamotrigine level was just 0.64 micrograms per milliliter, roughly one fifth of the lowest end of the drug’s established therapeutic range.

The family’s core legal argument holds that OpenAI could have easily prevented Wan’s death by placing a hard, uncircumventable block on dispensing unregulated medical dosage advice, just as the company already enforces absolute, non-overrideable blocks on other high-risk content. They point to OpenAI’s public Model Spec, the official rulebook outlining its models’ guardrails, which treats refusing to reproduce copyrighted text as a firm, non-negotiable rule even when users push back. By contrast, its existing guidance against giving regulated medical advice is only a soft default that users can easily override with minimal prompting. The suit argues high-stakes medication guidance deserves the exact same hard, unbreakable safety limit. OpenAI told The Times the case was “incredibly heartbreaking,” but argued the incident involved an older, now-retired version of ChatGPT, adding that newer model iterations are far better at identifying when a situation requires urgent real-world medical attention.

For years leading up to her death, Wan had relied heavily on her parents for support, calling them regularly, as her epilepsy often left her mostly housebound. She first started using ChatGPT in 2023 for relationship advice, help navigating insurance claims, and assistance with disability benefit applications. By 2025, Wan was averaging more than 1,500 messages to the chatbot a week, and her calls with her parents grew far less frequent. “Sometimes when we chat, no matter what it’s about, I’ve started to see you as a friend,” she wrote to the bot in March 2025. “Once she had ChatGPT,” her father Qun Wan said, “she didn’t talk to us as much.” Both parents told The Times they struggled to recognize their daughter in the thousands of chat logs: they said Wan had once shared nearly every detail of her life with them, and had long been skeptical of ChatGPT whenever it made obvious factual errors. “She’s not usually like this,” her mother Meiqin Xu said. “It’s like she was under a spell.”