STUDY: ROBOT DOCTORS WRONG HALF THE TIME, STILL ACCEPTING PATIENTS

Researchers from the US, Canada and UK asked ChatGPT, Gemini, Meta AI, Grok and DeepSeek ten questions in each of five health categories. About half the answers were rated problematic and nearly 20% highly problematic, worst for open-ended questions, nutrition and stem cells. The chatbots almost never declined to answer.

AI DOCTOR BOOTH AT THE MALL CLOSES WITHOUT WARNING, PATIENTS ADVISED TO EMAIL

Forward, a San Francisco health startup that raised about $400 million, shut down effective immediately in November 2024. Its CarePods were AI-powered self-service health kiosks placed in malls, gyms and offices, and its closure cost nearly 200 people their jobs.

SEARCH ENGINE RECOMMENDS GLUE ON PIZZA, ROCKS FOR BREAKFAST

In May 2024 Google's AI Overviews went viral for answers such as adding glue to pizza so the cheese sticks and eating rocks for nutrients. Google said the answers came from a joke forum post and a satirical site, and limited forum sources and AI answers for some health queries.

EATING DISORDER HELPLINE REPLACES HUMANS WITH BOT. BOT RECOMMENDS DIETING

The National Eating Disorders Association pulled its chatbot Tessa in June 2023 after a consultant shared screenshots of it recommending weight loss and calorie limits. NEDA had closed its human-staffed helpline the month before, and the bot's operator added generative AI to what was meant to be a scripted tool.