She sought medical advice from ChatGPT and died a few weeks later.

Bowen Wan, a 31-year-old software engineer, suffered from epileptic seizures. Her condition cost her her high-powered job at Amazon. She then turned to ChatGPT to adjust her medication and try to regain her speech, which had been impaired by the drugs. She died a few weeks later.

February 2026. Mid-conversation, Bowen Wan collapses. On the other end of the line, her father suspects another epileptic seizure. After a long moment, she regains consciousness and assures her father that she will call him back after resting. The wait drags on, and her parents, who live in China, video call one of their daughter's friends, who lives in Seattle. It is he who discovers Bowen's lifeless body.

“It would have been professional misconduct on my part to ask to stop medication when it was clearly working.”

A few weeks before her death, she had consulted artificial intelligence to adjust her medication from 400 to 300 milligrams per day, and then to 200 milligrams. A few months earlier, she had already reduced the dose of her antidepressants and planned to adopt a similar approach for her antiepileptics. This choice was driven by the desire to present a positive image during future job interviews, without being hampered by cognitive and speech problems caused by the medication.

According to her father, the chatbot encouraged her to reduce her dosage. It reportedly advised her to consult a doctor before praising her for not "blindly following" a medical prescription, the Daily Mail reports. The chatbot suggested that a doctor could prescribe the dose they deemed appropriate, but that she could begin adjusting her treatment herself.

Her neurologist, Dr. Nicholas Poolos, asserted that he would “under no circumstances” have recommended reducing the dosage of anti-epileptic drugs for a patient like Bowen Wan. “It would have been malpractice on my part to ask her to stop medication when it was clearly working,” he told The New York Times.

After following the advice provided by the AI, Bowen's health deteriorated, and she repeatedly turned to the chatbot to describe her symptoms. She mentioned episodes of confusion, memory loss, difficulty reading, and a strange sensation often associated with temporal lobe seizures. The complaint alleges that the chatbot consistently downplayed the severity of her symptoms. After a particularly alarming episode, she asked it if she should be concerned. The response was negative, according to the complaint.

OpenAI responds

A few days before her death, she confided in the AI that she felt exhausted and wondered if it wouldn't be better to resume her treatment. Her idea was not reassured by the artificial intelligence, which encouraged her to "persevere." The last message was sent three days before her death.

The Wan family has filed a lawsuit against OpenAI, which operates ChatGPT. A company spokesperson, interviewed by the New York Times, expressed regret over the “particularly heartbreaking” tragedy and clarified that Bowen Wan was using GPT-4o, an older version of the chatbot, now better programmed to handle high-risk situations. OpenAI also emphasized that ChatGPT is not a healthcare professional and reiterated the importance of consulting a doctor.