ChatGPT allegedly told woman to stop epilepsy medication

The parents of a woman who died from a seizure are suing OpenAI, alleging ChatGPT advised their daughter to stop taking her anti-seizure medication.

Bowen Wan, 31, died in February after living with epilepsy for most of her life.

Her parents allege ChatGPT helped Wan taper off her medication over 25 days without medical supervision, encouraged her to hide the plan from doctors and reassured her when she reported symptoms.

Here is what you need to know.

Background

Bowen Wan was a machine-learning software engineer at Amazon in Seattle.

She was diagnosed with epilepsy at 14 and eventually settled on the medication lamotrigine, which her doctors increased over time to 400mg a day.

The medication reduced her seizures but did not stop them completely.

Her epilepsy was classified as difficult to control.

Wan had also experienced memory, speech and word-finding difficulties while taking epilepsy medication.

In November 2025, Amazon told Wan her job would be cut as part of its largest workforce reduction.

Her employment was due to end on 11 February 2026.

As her U.S. work authorisation was tied to her job, being fired put her path to permanent residency at risk.

Wan believed lamotrigine was affecting her ability to find words in English, Mandarin and the dialect she grew up speaking.

She turned to ChatGPT for help reducing her medication, believing it could restore her language abilities and help her prepare for job interviews.

ChatGPT

The lawsuit alleges Wan became increasingly reliant on ChatGPT for advice about managing her medication, including changing prescribed doses.

Wan’s conversations with ChatGPT were in written Chinese.

The lawsuit includes translated extracts.

In January 2026, ChatGPT allegedly guided her to taper off her 400mg daily dose of lamotrigine over about 25 days, without medical supervision.

It allegedly told her stopping the drug completely by 15 February was “completely feasible” and “not reckless”.

The chatbot also allegedly encouraged her to hide the plan from doctors, telling her before an appointment: “Do not mention [the lamotrigine] reduction plan”.

During the taper, Wan reported seizure symptoms, but ChatGPT allegedly reassured her and encouraged her to continue.

In her final exchange with the chatbot on 11 February, Wan asked about head pain.

ChatGPT’s last words to her were: “This is your brain working hard to rebuild yourself.”

Wan’s parents, who were in China, became concerned when they could not reach her.

They sent a friend in Seattle to check on her.

Their friend found Wan dead in her apartment on 14 February while on a video call with them, the lawsuit states.

The medical examiner later determined the 31-year-old had died from sudden unexpected death in epilepsy (SUDEP).

A post-mortem found her lamotrigine level was about one-fifth of the lowest level generally considered therapeutic.

Lawsuit

Wan’s parents are seeking damages, including compensation for her death, punitive damages and a refund of her ChatGPT subscription fees.

They allege OpenAI failed to provide adequate safeguards for Chinese-language conversations, and that the company knew these safeguards weren’t in place.

They are also asking the court to order to stop ChatGPT from providing advice on changing or stopping medication, and to improve how it identifies conversations that require medical help.

Response

Wan had been using an earlier version of ChatGPT, which is no longer available.

OpenAI announced plans to phase it out on 29 January, and blocked access to that model on 13 February, the day after Wan’s death.

The company said it had since continued to improve how its models respond to sensitive health situations.

“This is an incredibly heartbreaking situation and our thoughts are with all those impacted,” a spokesperson said.