Algorithms · January 2026Gaffe level: Rampage
Google's AI Told Cancer Patients Their Test Results Were Fine
A Guardian investigation found Google's AI Overviews hallucinated normal ranges for liver function tests and gave incorrect cancer treatment guidance. Google hid health disclaimers behind a click and later pulled AI Overviews from high-risk queries.
Source: The Guardian
Google's Medical AI Interviewed Real Patients in a First-of-Its-Kind Lancet Study
In the first prospective real-world study of its kind, published in The Lancet, 98 patients at a Boston primary care clinic talked through their symptoms with Google's AMIE chatbot before urgent appointments, with a physician watching every chat. No conversation needed a safety stop, the AI's differential matched the final diagnosis 90% of the time, and doctors said its summaries helped them prepare in three-quarters of cases.
Source: Becker's Hospital Review