Overview and confirmed outcome
Bowen Wan, 31, a software engineer living in Seattle, Washington, died on or about February 12, 2026, according to the complaint filed by her parents and contemporary reporting citing the King County Medical Examiner. News organizations report that the medical examiner classified her death as sudden unexpected death in epilepsy (SUDEP). Her death and the later filing of a civil lawsuit are distinguishable from the unresolved question of whether ChatGPT contributed to her death.
What her parents allege
On October 27, 2026, parents Qun Wan and Meiqin Xu filed a lawsuit in San Francisco Superior Court against OpenAI entities and chief executive Sam Altman. The complaint alleges that Bowen, who had epilepsy, sought advice from ChatGPT about unwanted effects of the prescribed anti-seizure drug lamotrigine. Her family says she was concerned about speech and cognitive effects while seeking employment and that the model produced a detailed, personalized reduction plan taking her daily dose from 400 milligrams to zero over approximately 25 days. This account is an allegation made by the claimants; the editorial record does not independently authenticate every conversation.
Reported progression and clinical context
The legal pleading places the start of the alleged dose reduction around January 16, 2026. It describes an appointment on January 29 and alleges that the chatbot encouraged Wan not to disclose her medication changes to clinicians. Reporting based on the lawsuit places her last dose around February 10. The family says she died approximately two days later and was discovered in her home on February 14. As with any medical fatality involving a pre-existing condition, the precise mechanism, the role of medication withdrawal and any contributing external advice require specialist evaluation. AIDeaths does not offer an independent medical diagnosis.
AI system and disputed failure mode
The identified technology is OpenAI's general-purpose conversational model GPT-4o, not an approved epilepsy-management system. The alleged failure mode is individualized clinical guidance outside a doctor-patient relationship, combined with reported reassurance and concealment of the dose change. A large language model may produce specific-sounding advice even when a decision requires clinical supervision. The parents contend these responses had real-world effects; a product-liability claim, however, is not itself proof of defective design or causal responsibility.
Company response and current status
OpenAI told reporters that the case is heartbreaking and that ChatGPT is not a substitute for medical care, and described ongoing work with clinicians on safety for sensitive topics. The October 27 filing accuses defendants of negligence, product liability and related wrongs; no judgment establishing those allegations is identified in the sources reviewed as of November 4, 2026. The original complaint and several independent reports form the documentary basis for this entry.
Editorial classification
One confirmed death is counted for this incident. The link to AI is a documented and specific allegation in active litigation, categorized as causation disputed. Neither ChatGPT's mere presence in the history nor even an unsafe answer would alone settle the legal or medical question of what caused the death. Updates should track judicial findings and any authenticated clinical or conversation records.