Engineer, 31, dies after befriending ChatGPT bot which convinced her to change her medication dose

A young and gifted engineer died after a ChatGPT bot she trusted told her to reduce her anti-seizure medication, according to a lawsuit filed by the woman's parents.

Bowen Wan, 31, was diagnosed with epilepsy at the age of 14, but she managed the debilitating condition throughout her life with the help of medication.

Wan was a prodigy from a rural town in China who was admitted to elite schools in the country before moving to the US halfway through college and graduating from Arizona State University with a computer science degree. 

She later moved to Seattle and got a job as a machine-learning software engineer at Amazon. The young woman kept in touch with her parents on the other side of the Pacific through phone and video calls. 

But Wan's parents, Qun Wan and Meiqin Xu, became concerned in February after they had not heard from their daughter in two days. 

They asked a friend in Seattle to check on Wan in her apartment. The parents were on a video call with the friend when he entered the apartment and found Wan dead, lying facedown in her bed, according to the lawsuit. 

The young woman was in her pajamas, with reddish-brown blood stains on her pillow and bedding. Pills of the anti-seizure medication lamotrigine that kept her alive were strewn about the floor. 

An autopsy determined that she had died from Sudden Unexpected Death in Epilepsy. Wan's parents were shocked. Their daughter had successfully managed her condition for half her life, so how could this have happened, they wondered.

Bowen Wan, 31, died after ChatGPT told her to reduce her anti-seizure medication, according to a lawsuit filed by her parents 

Wan's parents asked a friend in Seattle to check on her after they had not heard from her for two days. The parents were on a video call with the friend found Wan dead

As it turned out, Wan had been taking damaging medical advice from ChatGPT after developing a relationship with the AI language model for about three years, according to the lawsuit. 

Wan's use of ChatGPT began fairly modestly. She used it for research and sent an average of around 1,300 messages a month in 2023, the lawsuit said. 

Over time, she began trusting the chatbot more, sending it 10,761 messages in September 2025, and using it for advice on a range of issues, according to the lawsuit. She would even consult the AI as a 'friend.' 

'I don’t know why, but sometimes when we chat, no matter what it’s about, I’ve started to see you as a friend,' Wan messaged ChatGPT in March 2025, the New York Times reported. 

Wan paid $200 per month for ChatGPT, the highest-tier subscription for consumer users, because it allowed her to keep using the model she had connected to, GPT-4o.

GPT-4o has been accused of being dangerously agreeable and contributing to users ending their own lives. 

Several wrongful death lawsuits have been filed against OpenAI by the families of people who killed themselves after the GPT-4o model allegedly encouraged them to do so.

In August 2025, following a spate of those lawsuits, OpenAI replaced the model with a demonstrably safer one, GPT-5, according to the lawsuit. 

Wan built a relationship with a ChatGPT model called GPT-4o, which has been accused of being dangerously agreeable and contributing to users ending their own lives, the lawsuit said

Wan, pictured during her graduation from Arizona State University, was paying $200 per month for a ChatGPT subscription so she could keep using GPT-4o, the lawsuit said

But users objected, so the company allowed paying subscribers who 'preferred GPT-4o’s conversational style and warmth' to continue using the model, the lawsuit said. 

Wan was one of those users who preferred GPT-4o. In January 2026, she told the chatbot, 'If 4o is gone someday and only 5.1 is left, I’ll stop using ChatGPT,' according to the suit. 

In earlier instances, she expressed concern about her reliance on the model. 'In hard times I talk with GPT-4o too much,' she messaged the bot in April 2025, the lawsuit said.

It replied by affirming her use. 'In any case, the most important thing is that you keep enjoying talking with GPT-4o!' the chatbot told her, according to the lawsuit, which also described the bot as 'sycophantic.' 

In November 2025, Wan was notified by Amazon that her employment would be terminated in 90 days. Since she was in the US on a work visa, that meant she had just a few months to find a new job and remain in the country. 

Wan felt that the side effects of her anti-seizure medication made it difficult to think clearly and were causing her to lose proficiency in English, which would be a problem for job interviews. 

So, after having built trust in ChatGPT for years, she turned to the bot for advice. 

When she asked if it would be a good idea to wean off her anti-seizure medication, the AI said yes and provided Wan with a plan to taper her 400-milligram dose down to zero, according to the lawsuit. 

OpenAI replaced GPT-4o with a safer model after a spate of wrongful death lawsuits, but the company later restored access to 4o for paying subscribers, the lawsuit said

Over the course of the following 25 days, Wan followed the bot's plan but asked to accelerate the timeline on multiple occasions. Each time, ChatGPT consented, the lawsuit said. 

During the tapering period, Wan suffered an increase in seizures, as well as difficulty sleeping and confusion, the lawsuit said. 

But each time she came to ChatGPT with concerns, it reassured her that everything was fine and the plan was working as intended, according to the lawsuit. 

She began tapering on January 16, and she died on February 12.

OpenAI announced it would be completely retiring GPT-4o on January 29, when Wan was 13 days into her taper. The company fully removed access to the model on February 13, one day after Wan died. 

The lawsuit accused OpenAI and its CEO, Sam Altman, of negligence and wrongful death. It claimed that Altman and his company were aware of the dangers of GPT-4o, but rushed the model's release to be competitive with other AI companies.

OpenAI spokesman Drew Pusateri told the Daily Mail: 'This is an incredibly heartbreaking situation and our thoughts are with all those impacted. 

Working closely with clinicians and other experts, we have continued to strengthen how ChatGPT responds in sensitive and acute situations, while also continually improving how our latest models engage with users more broadly.' 

The lawsuit accused OpenAI CEO Sam Altman of rushing the release of GPT-4o despite being aware it was dangerous in order to compete with other AI companies

OpenAI also emphasized that GPT-4o is no longer available to users and that the company works with hundreds of physicians and experts to make its models safer.

The spokesperson added: 'While ChatGPT is not a substitute for medical care, we believe AI can play a role in helping people make sense of health information, navigate insurance and complex healthcare systems, and prepare for appointments with medical professionals.'

If you or someone you know needs help, please call or text the confidential 24/7 Suicide & Crisis Lifeline in the US on 988. There is also an online chat available at 988lifeline.org.