A woman died after ChatGPT convinced her to stop her epilepsy treatment

The parents of 31-year-old Bowen Wan have filed a lawsuit against OpenAI following the death of their daughter, who had stopped taking her epilepsy medication.

According to the plaintiffs, ChatGPT effectively provided illegal medical advice, encouraged the woman to conceal her actions from doctors, and suggested a dangerous plan to gradually wean her off life-saving medications. The parents claim that this led to her death.

The official cause of death was sudden unexpected death in epilepsy (SUDEP)—a phenomenon that often occurs after a severe seizure. However, an analysis of the deceased’s phone messages revealed that she had sharply reduced her dose and completely stopped taking lamotrigine following direct instructions from artificial intelligence.

Bowen Wan moved from China to the U.S. and worked as a software engineer at Amazon. She had a severe form of epilepsy and was taking medication. However, she complained of side effects, which she said her doctors ignored. According to her, the medication caused her speech to become unnaturally choppy and her thinking to slow down.

Later, she learned she was among those affected by a round of layoffs at Amazon. She was afraid of losing her visa and was eager to find a new job. To present herself in the best possible light during job interviews, Bowen Wan began considering stopping the medication and turned to ChatGPT—which endorsed her plan to reduce the dosage on her own from 400 mg to 0 mg in just a few weeks.

The chatbot not only prescribed a dosage reduction plan but also actively undermined her trust in medical professionals. When Wan expressed doubts, the system assured her that doctors’ recommendations were supposedly driven by fear of risks rather than concern for her well-being, and praised the patient for her decision to act in secret.

According to the investigation, the young woman expressed concern about her deteriorating condition at least 158 times. However, the algorithm interpreted dangerous seizures, loss of consciousness, and a feeling of pressure in her head as positive phenomena—allegedly, her brain was “switching to low-power mode” or “working on self-healing.” The woman’s body was found in her bed a few days after her last message to the algorithm.

OpenAI spokesperson Drew Pusateri called the incident tragic and noted that the GPT-4o model was used in the conversations, which the company removed from public access immediately after the incident. The company stated that newer versions assess life-threatening risks more strictly.

Source: The New York Times.

In Russia, a woman was fined for a fake photo of a bear created by ChatGPT — Reuters.

ChatGPT has introduced a four-digit PIN code to lock chats — Engadget.

OpenAI has launched a separate version of ChatGPT for teenagers.