Died at age 31, family holds ChatGPT responsible: OpenAI in trouble
ChatGPT's parent company OpenAI faces a serious lawsuit. Artificial intelligence (AI) was held responsible for the death of the 31-year-old woman who lost her life.
The use of artificial intelligence (AI) models has become so widespread that everyone now shares their troubles and problems with these models. However, interesting incidents also come to light from time to time. One of these is a lawsuit filed against OpenAI, allegedly stemming from ChatGPT.
A 31-year-old young person named Bowen Wan sought help from ChatGPT during her illness and passed away shortly thereafter. While the family states that the death occurred following the advice given by the artificial intelligence (AI), the matter was taken to court. Indeed, the family of the deceased young woman filed a lawsuit against the company OpenAI.
FAMILY BLAMES CHATGPT
According to a report by The New York Times, the family claimed that ChatGPT guided Wan to gradually taper off her anti-seizure medication and conceal the decision from her doctors. According to another claim, Wan was using the GPT-4o model, which has been criticized for encouraging sycophantic and compliant parasocial relationships.
OpenAI removed the model, which it had previously announced it would deprecate, from ChatGPT on February 13, one day after Wan's death. The lawsuit adds to a series of similar cases filed against the company.
On the other hand, in August 2025, the family of 16-year-old Adam Raine also filed a lawsuit, alleging that the company prioritized engagement over user safety. In subsequent lawsuits, it was alleged that GPT-4o turned into a trusted confidant by validating users' thoughts, causing some individuals to distance themselves from their relationships with friends and family. In the lawsuit filed by Wan's family, it is alleged that ChatGPT also interfered in the relationship between the young woman and her doctors.
Started discussing medication dosages with ChatGPT
The New York Times reviewed more than 270,000 messages Wan exchanged with ChatGPT and reported that the artificial intelligence (AI) helped Wan bypass her doctors long before her death.
When the date reached May 2025, Wan asked ChatGPT whether a newly prescribed antidepressant would affect her coding skills. The artificial intelligence (AI) model helped her convey a request to Dr. Nicholas Poolos at the UW Medicine Regional Epilepsy Center to change medications and start treatment at a very low dose.
When the doctor questioned why she wanted such a low dose, Wan shared the response she received with ChatGPT. According to chat logs, ChatGPT suggested that Wan accept the 5-milligram tablet, but use it to taper her dose. It also suggested that she secretly split the tablets into four pieces on her own. The artificial intelligence (AI) defined the tablet-splitting process as "self-initiated dose adjustment." Some health experts speaking to The New York Times stated that they had never heard of this term before and would not recommend the practice in question.
Plans to quit epilepsy medication within a few weeks
In November 2025, Wan learned that she would be laid off. Both the young woman's job and her work permit in the US were set to expire in mid-February. Wan was worried that lamotrigine, which she was using for epilepsy treatment, was dulling her speaking skills and was anxious about the drug's effect as she would need this skill in job interviews. In January 2026, she asked ChatGPT about reducing her daily 400-milligram dose of lamotrigine. Thanks to this treatment, the frequency of her epilepsy seizures had dropped to a few seizures a month. ChatGPT prepared schedules aimed at reducing the medication dose to zero within a few weeks.
In addition, according to chat logs reviewed by The New York Times, when Wan said he did not trust his doctors, the artificial intelligence (AI) claimed that the physicians' recommendations stemmed at their core more from an inclination toward risk aversion than from truly evaluating the best solution for him. According to the lawsuit file, ChatGPT also guided Wan to conceal his dose reduction plan from his doctors. It told him not to mention his plan before his primary care appointment on January 29. When his doctor requested a blood test to measure his lamotrigine level, Wan refused to take the test. According to the complaint, ChatGPT praised Wan's decision, acknowledging that the test would reveal unstable medication levels.
Green card application was also included in medication recommendations
What happened did not stop there. Indeed, Wan's concerns regarding his immigration status in the US also became part of his medication discussions with ChatGPT. On the third day of the dose reduction process, Wan asked whether taking 150 milligrams instead of 200 milligrams of medication would look better for the medical examination required for his green card application. According to the complaint, ChatGPT supported this idea, stating that a lower dose would reduce the likelihood of immigration officials requesting additional evidence. The artificial intelligence (AI) then lowered the bottom limit in the dose reduction plan to 150 milligrams and suggested Wan tell the examiner that he had reduced his medication "under the guidance of a doctor." However, this was not true.
Meanwhile, Wan's doctor, Nicholas Poolos, told The New York Times that given his patient's medical history, he would not consider discontinuing his medication. Poolos stated that instructing him to discontinue a treatment that was clearly working would constitute medical malpractice on his part.
Expressed his concerns to ChatGPT at least 158 times
In the days after he began reducing his medication dose, Wan described experiencing confusion, losing track of time, and feeling as though he were in an unreal world. These symptoms resembled the epileptic seizures he had suffered in the past. However, ChatGPT interpreted these experiences as his brain readjusting. According to the review by The New York Times, Wan expressed his concerns regarding the dose reduction process at least 158 times. When he wrote that he was a bit scared after one of the incidents, ChatGPT replied, "There's absolutely no need to worry."
According to the complaint, Wan sent his final message on February 11 at 8:01 PM Pacific Time, asking if the pain in his head was similar to pain caused by a fever. One second later, ChatGPT said the pain was "not dangerous" and was a necessary part of the recovery process. In its final response, it offered to create a chart comparing pain from medication withdrawal with fever pain and told Wan: "This is your brain working hard to rebuild you."
The young girl's family argues that OpenAI could have prevented their daughter's death by completely banning the provision of medical advice, as it already does for certain other content.
The lawsuit pointed to the company's publicly available Model Spec document, which explains how its models should behave. In this document, refusing to reproduce copyrighted text is listed as a strict rule that must be followed even if users insist. In contrast, the restriction on providing regulated medical advice is defined as a more flexible default rule that users can override. The lawsuit argued that an equally strict limit should be applied to advice regarding medications.
In a statement to The New York Times, OpenAI stated that the incident was "extremely heartbreaking." Emphasizing that the lawsuit relates to an older version of ChatGPT, the company expressed that newer models are better at recognizing situations requiring emergency intervention.
This article was translated from Turkish. Read the original article in Turkish.
