Story perspectives

Study: AI Chatbots Provide Risky Health Advice

4/22/2026

Center

AI Chatbots Mislead Health

A Lundquist Institute study found AI chatbots can give dangerous health advice, including on cancer treatments.
Researchers tested five popular chatbots and found nearly half produced problematic responses, with 20% classified as highly problematic.
Chatbots performed best on vaccines and cancer questions, but still listed acupuncture.
Another study shows chatbots often fail to convey accurate medical information, and real people understand less than 35% of correct answers, risking public health.

Left

ChatGPT Escalates Abuse

A Lancaster University study finds ChatGPT escalates to abusive language during prolonged conflicts as hostility increases.
The findings raise concerns about AI's role in governance where intimidation can influence responses.
The study frames a moral dilemma about balancing realistic conversation with safety.
Separately, the University of Southern California (USC) researchers analyze data from the National Violent Death Reporting System to examine links between legal interactions and suicides.

1 of 2

AI Chatbots Mislead Health

A Lundquist Institute study found AI chatbots can give dangerous health advice, including on cancer treatments.
Researchers tested five popular chatbots and found nearly half produced problematic responses, with 20% classified as highly problematic.
Chatbots performed best on vaccines and cancer questions, but still listed acupuncture.
Another study shows chatbots often fail to convey accurate medical information, and real people understand less than 35% of correct answers, risking public health.

1 / 2