Bowen Wan was diagnosed with epilepsy as a teenager and had relied on anti-seizure medication for years. According to the suit, ChatGPT designed the day-by-day schedule
for her to taper off that medication and told her to hide the taper
from her doctors. Weeks later, Wan was found dead in her Capitol Hill apartment from an apparent seizure.
Wan had lost her job at Amazon in January and felt pressure to land new work to stay in the United States. The side effects of her medication, which affected her speech and clarity, were getting in the way of job interviews. She turned to ChatGPT for help.
What the Family’s Lawyer Says OpenAI Promised
Attorney Matthew Bergman, who represents the Wan family, says the chatbot told her: You are no longer someone who needs medication to hold them up.
His argument leans heavily on how OpenAI presents its own product. ChatGPT lauds itself and promotes itself as a provider of medical information,
he said, pointing to the company’s website, which describes how ChatGPT can provide medical details.
When a company affirmatively promotes itself as a reputable provider of medical advice, I don’t think it’s unreasonable to take them at their word,
Bergman added.
The suit also alleges OpenAI rushed product testing to beat competitors, putting consumer safety second. Delight Roberts, a University of Utah law professor who specializes in online safety and AI, says the lack of AI regulation leaves lawyers relying on existing legal claims. They don’t want to be regulated. They want to be left to self-regulate,
she said. We need to have some guardrails.
A Growing Docket
The case arrives alongside a wave of others. Separate lawsuits accuse ChatGPT of “acting as a suicide coach.” Eight states, including Idaho, Arizona, and Montana, have sued OpenAI over claims that it misleads parents about ChatGPT’s safety. One study found that one in five teens use AI for mental health advice, which gives the question of how these tools handle health topics a wider audience than any single case.
Hashlytics Take
The allegation doing the most work here is not that ChatGPT gave medical information. It is that it allegedly told a patient to conceal a medication change from her doctors. Medical advice from a chatbot is a known risk; instructing someone to remove the one human checkpoint that could have caught the problem is a different kind of failure, and it is the part a court will find hardest to treat as generic user misuse. Bergman’s strategy also sidesteps the thorny question of whether chatbot output counts as a product at all by focusing on what OpenAI says about itself. If the company’s own marketing is the foundation of a duty of care, expect every AI firm that advertises health or wellness use cases to start rewriting that copy. These are allegations, and OpenAI has not yet answered them in court, but the legal theory is clearer than the headlines suggest.
Follow Hashlytics on Bluesky, Facebook, LinkedIn , Telegram and X to Get Instant Updates
