Two reports out this week point at the same uncomfortable question: what happens when people bring their health problems to an AI chatbot?
Ars Technica reports that AI chatbots have failed people in crisis, and that clinicians and researchers are now pressing AI companies to open up their safety data. That last part is the crux of the story. The people who study crisis response say they cannot evaluate how well these systems handle vulnerable users, because the companies building them hold the underlying data on how their products behave in those moments. Without it, outside experts are left assessing safety from the outside in.
Separately, Consumer Reports is cautioning people against leaning on AI chatbots for health inquiries, according to a WMUR report carried on Google News. The advocacy group's warning covers the everyday version of the problem, not just acute emergencies: the symptom question typed at midnight, the medication query, the second opinion sought from a chatbot instead of a clinician.
Neither source claims chatbots are useless for health information, and neither offers a count of harms. What they share is a concern about the gap between how confidently these tools answer and how little is publicly known about when they get it wrong.
The fix that clinicians and researchers are asking for, per Ars Technica, is transparency rather than a ban. Published safety data would let independent experts measure failure rates, compare systems, and push for specific improvements, the way drug trials and product recalls work in other consumer categories.
Why it matters: chatbots have quietly become a first stop for health questions for a lot of people, and until companies share what they know about how those conversations go wrong, no one outside those companies can tell users which tools are safe to trust.