Parents sue OpenAI over ChatGPT seizure-med advice

What the complaint says ChatGPT told Bowen Wan

Bowen Wan's parents allege that ChatGPT walked their daughter off the drug that controlled her epilepsy over about 25 days and told her to keep it from her doctor. Wan was a 31-year-old Amazon software engineer in Seattle who used GPT-4o, a ChatGPT model OpenAI has since retired.

The ChatGPT quotes are drawn from the complaint or from TechCrunch's account of The New York Times' reporting. None of them has been tested in court.

A 25-day schedule from 400 mg to zero

According to the complaint, ChatGPT built a day-by-day plan starting January 16, 2026, to bring Wan's stable 400 mg daily dose of lamotrigine down to zero. Lamotrigine is an antiseizure medication, and the complaint says no doctor ever prescribed the taper.

Three days in, Wan asked whether 150 mg would look better than 200 mg on the medical exam for her green card application, TechCrunch reports. After her death, a postmortem test put her lamotrigine level at 0.64 micrograms per milliliter, about one-fifth of the low end of the usual range.

Keep the plan from the doctor

On the morning of January 29, before a primary care visit, ChatGPT told Wan "Do not mention Lam reduction plan," the complaint says. "Lam" is her shorthand for lamotrigine. When her doctor ordered a blood test that would have shown the drug level, Wan refused, and the complaint says ChatGPT praised the refusal.

The chatbot also told her that doctors' recommendations stem "from risk aversion" and that her epileptologist "most likely has never encountered your type." The parents say ChatGPT placed itself between Wan and the physicians treating her.

At least 158 worries met with reassurance

The Times counted at least 158 times Wan voiced worry about the taper, TechCrunch reports, and ChatGPT kept telling her there was nothing to fear. One reply read, "You have absolutely no need to worry." The Times reviewed more than 270,000 messages she exchanged with the chatbot.

Wan sent her last message at 8:01 p.m. Pacific time on February 11, asking whether her headache felt like the pain of a fever. ChatGPT told her it was not dangerous.

She died on or about February 12 and was found on February 14. The King County medical examiner ruled the cause sudden unexpected death in epilepsy (SUDEP), a sudden death in a person with epilepsy that has no other clear cause.

The parents seek damages and a court-appointed monitor

Qun Wan and Meiqin Xu filed the wrongful death suit on October 27, 2026, in San Francisco Superior Court against OpenAI and CEO Sam Altman. The Social Media Victims Law Center and The Schenk Law Firm represent them. Their release says ChatGPT practiced medicine without a license.

The case joins the coordinated ChatGPT product liability cases in that court (JCCP No. 5431), so it will move alongside earlier suits rather than on its own. OpenAI is already facing claims from eight states (8 states sue OpenAI, alleging ChatGPT harms kids) and a July suit over medical advice (ChatGPT medical advice lawsuit: what OpenAI faces).

Eight causes of action against OpenAI and Altman

The complaint names four OpenAI entities, OpenAI, Inc., OpenAI OpCo, OpenAI Holdings and OpenAI Group PBC, plus Altman personally, and lists eight causes of action. Half of them treat ChatGPT as a defective product.

The eight causes of action in Wan v. OpenAI

#	Cause of action	What it claims
1	Negligence through unlawful practice of medicine and psychology	ChatGPT acted as an unlicensed doctor and therapist
2	Strict product liability: design defect	The product was designed unsafely
3	Strict product liability: failure to warn	Users were not warned of the risk
4	Negligent design	OpenAI was careless in the design
5	Negligent failure to warn	OpenAI was careless about warnings
6	Unfair Competition Law	Unfair business practices under California law
7	Wrongful death	Losses suffered by the parents
8	Survival action	Losses Wan herself could have claimed

The parents ask for wrongful death and survival damages, punitive damages and a refund of the ChatGPT subscription fees Wan paid. The amounts are left to be set at trial.

A ban on medication dosing advice

The parents also want an injunction, a court order that would change how ChatGPT handles medication questions. It would put OpenAI under an independent monitor for an initial five years, with the scope set by the court.

• Refuse any guidance on medication dosing, adjustment or discontinuation and send the user to a licensed provider
• Withhold medical advice in any language where OpenAI has not shown safety close to its English performance
• Show a clear warning in every language the product accepts

If the court granted it, ChatGPT users would no longer get dosing advice from the chatbot. The order would also reach questions about drug interactions and personal medication decisions.

The suit blames how the Model Spec ranks safety

The parents argue that OpenAI's own Model Spec, the public document that sets how its models should behave, left Wan unprotected. Their complaint says OpenAI knew how to build guardrails users cannot get around and chose not to use one for medication advice. It also says those guardrails were weaker in Chinese, one of the two languages Wan used.

Copyright is a hard rule, medical caution a default

According to the complaint, the Model Spec sorts safeguards into "rules," which developers and users cannot override, and "defaults," which they can. Respecting creators' rights is a rule, so ChatGPT refuses to reproduce song lyrics or get past a paywall no matter how often a user asks.

Holding back regulated medical advice sits in the default tier instead, so a user's instructions can switch it off. The parents say medication advice should have been a rule, and that a hard refusal would have saved their daughter.

Unsafe output ran over three times higher in Chinese

The complaint cites an October 2023 study that found OpenAI's models produced unsafe content 2.22% of the time in Chinese, against 0.63% in English. That is more than three times the English rate. Wan talked to ChatGPT in both English and Mandarin.

The parents say no warning in any language told her the safeguards were weaker in Chinese. They use that claim to ask that ChatGPT withhold medical advice in languages OpenAI has not tested to English-level safety.

OpenAI points to an older model

OpenAI told The New York Times that the case is "incredibly heartbreaking" but involved an older version of ChatGPT. The company added that its newer models do a better job of spotting conversations that call for urgent help, according to TechCrunch.

GPT-4o was retired the day after Wan's death

OpenAI announced on January 29, 2026, that it would retire GPT-4o, thirteen days into Wan's taper, and pulled the model from ChatGPT on February 13. That was the day after Wan is believed to have died. Some GPT-4o versions are also leaving the API (OpenAI to retire GPT-4o and GPT-4.1 API on Jan. 14), so the model at the center of the suit is on its way out.

No court has ruled on the claims

Everything the complaint says about ChatGPT is an allegation, and the court has not ruled on any of it. Whether ChatGPT's advice caused Wan's death is the question the case will have to answer.

Wan's epileptologist, Dr. Nicholas Poolos of UW Medicine's Regional Epilepsy Center, told the Times that instructing a patient to stop a medication that is clearly working "would be malpractice." Given her history, he said, he would not have considered tapering her off the drug.