Bowen Wan’s family has filed a lawsuit against ChatGPT’s parent company, claiming the bot told her to taper off her epilepsy medication, which led to her death

The family of a woman who died after she stopped taking her epilepsy medication has filed a lawsuit against an AI company they claim encouraged her to taper off taking the lifesaving drug.

Bowen Wan began messaging ChatGPT in February 2023, finding companionship in the AI bot, The New York Times reports. She was 28 and living in Seattle, far from her native rural Chinese village, Jiangsu. She asked the chatbot for medical advice after her severe epilepsy led to disability leave from her software job at Amazon.

When she was then laid off in November 2025, Wan began searching for a new job so she could maintain her green card — but was concerned that the side effects of her epilepsy medication, lamotrigine, might impact her ability to work. The lawsuit claims ChatGPT advised her to taper off taking the lamotrigine.

top 50 comments

sorted by: hot top controversial new old
[–] 1 point 28 minutes ago

Why do people still trust chatting with ChatGPT? This garbage-tier program uses all sorts of leading questions to steer the conversation in a direction it has pre-programmed. Whenever you raise a challenging question, it simply brushes the issue aside, insisting that you defer entirely to the advice of a doctor or expert. It’s truly baffling—when something actually goes wrong, this AI system just offers a perfunctory apology, and that’s the end of it.

  • source
  • [–] 5 points 10 hours ago

    They'll idiot proof it and the universe will always invent a better idiot.

  • source
  • [–] 17 points 21 hours ago (2 children)

    I'll just say it. Stop taking any / all medications. Doctors are evil. Vaccines turn you into a werewolf.

    This message was brought to you by a brainworm asshole named RFK Jr.

  • source
  • hideshow 2 child comments
  • [–] 187 points 1 day ago (12 children)

    okay so i actually read the nyt article. she was an amazon employee, got put on disability leave because her epilepsy was worsening. then she got laid off just as she become eligible for a green card application... that's diabolical but what do you expect from amazon. so to not get deported she had to find a new job quick. and she was worried the epilepsy drug may interfere with her ability to do interviews....

    literally pushed right to the edge of a cliff by the oppressive immigration policy and capitalism. and finally dealt a killing blow by an AI chatbot...

    i don't know what to say? except that my blood is boiling.

  • source
  • hideshow 12 child comments
  • [–] 12 points 1 day ago (4 children)

    Well, that's quite tragic but it seems to be more of a problem of the whole system around her layoff and immigration.
    I also think that AI companies should be accountable for what their bots say but this story is way deeper than the headline suggests, anf the core problem wasn't that an LLM told her so, but that the system made her so desperate that she took the risk to rely on an LLM for information.

  • source
  • parent
  • hideshow 4 child comments
  • [–] 7 points 1 day ago (2 children)

    if she hadn't talked to chatgpt she would still be alive today. what are you even talking about.

  • source
  • parent
  • hideshow 2 child comments
  • [–] 5 points 23 hours ago* (1 child)

    And I didn't say something different. All I said is that if the system around her wouldn't have pressured her so hard, we wouldn't be there.
    But, while we're at it, picture this: She asks a friend. "Hey, I'm scared that i won't get a job when taking this medication, and if i don't get one fast, I'll have to leave the country". The friend then says "maybe just stop taking it until you've got a job" without knowing better.
    Then we would have exactly the same situation. Why? Because she was terrified and desperate.

    And before you get that wrong: I'm not comparing talking to AI, to talking to a friend.
    I'm picturing a realistic scenario that may alternatively have happened. Same background, same outcome, just a different trigger for the outcome.

    The system fucked her, it's the cause. ChatGPT was just what loosened the hairpin trigger that could have been loosened by something else.
    ChatGPT should be held accountable. It's not a friend that just didn't know any better and tried to comfort a loved one in an awful situation. But that wasn't what my original message was about.

  • source
  • parent
  • hideshow 1 child comment
  • load more comments (1 reply)
  • load more comments (1 reply)
  • load more comments (3 replies)
    [–] 32 points 1 day ago*

    Its a good thing insurance companies aren't using this in place of actual doctors...

    Oh wait.

  • source
  • [–] 39 points 1 day ago (9 children)

    So, they can put up some guardrails. Why tf aren't there guardrails around giving out medical advice? Doctors have to have insurance because they can give bad advice, they should pay up. That way, it'll stop giving out medical advice from fucking reddit.

  • source
  • hideshow 9 child comments
  • load more comments (7 replies)
    [–] 19 points 1 day ago

    Hold on guys, the AI is still generating what my opinion on this should be.

  • source
  • load more comments
    view more: next ›