LLM drug recommendations converge on single products, averaging 97.5% of first picks
EricTopol · x · 2026-10-26
- A Nature Health study finds five LLMs show extreme product-name convergence when recommending FDA-approved drugs with the same active ingredient: the most-recommended product averaged 97.5% of first-choice picks across models and disease–drug pairs.
- Scale: 177,120 synthetic clinical vignettes spanning 123 disease–drug pairs and 1,440 sociodemographic profiles.
- Key finding: LLM recommendations diverge from general medical practice and often point to higher-cost alternatives, potentially amplifying systematic biases affecting equity and cost.
More from Safety
- If an agent can rewrite itself, does your consent still mean anything? — tallmetommy · 2026-10-26
- 109 of 529 AI Agent Incidents Involved No Adversary, Arguing Red Teams Must Test Normal Operation — anacondainc · 2026-10-26
- Cryptographer Matthew Green: if your code looks vulnerable, fix it now — matthew_d_green · 2026-10-26
- Filtering secrets on the write path: why agent memory deserves more suspicion at ingest than retrieval — Classic-Being-7960 · 2026-10-26
- After OpenCode key leak: 5 practical rules for keeping keys away from agents — jacobholm_dev · 2026-10-26
- Formal verification spat: specs are inherently limited against real-world threats — davidmanheim · 2026-10-26