LLMs Fail Over 60% of Women's Health Scenarios in New Study, as AI Hallucinations Hit Courts
A study testing 13 leading LLMs on women's health found most couldn't identify basic clinical conditions — part of a broader pattern of AI failures now affecting legal proceedings and public discourse.
Thirteen leading large language models were tested on clinical scenarios involving women's health, and over 60% failed to identify the condition or clinical risk, according to @DrCatharineY. The post went viral, underscoring growing concern that LLMs are being deployed in healthcare contexts where their failure modes can cause real harm. The study didn't name which models performed worst, but the breadth of the failure — across 13 models, not one or two outliers — suggests a systematic gap in training data or evaluation rather than a single vendor's shortcoming.
Unlock the full briefing
Get every story in today's briefing, the full archive, and the daily AI intelligence brief.
All stories today
Full archive
Daily brief
Cancel anytime. Payments powered by Stripe.