The authors recruited 1,252 ChatGPT users from India, Nigeria, Brazil, and Pakistan who donated their full conversation histories (202,590 conversations, December 2022–February 2026) with self-reported age and gender.
ChatGPT's health answers are high-risk for one in five real users
The first safety evaluation of real health conversations with ChatGPT in four Global South countries reveals that context-aware scoring increases risk ratings and uncovers hidden harms.
Independent
Shreyasi Roy Chowdhury · Kiran Garimella
Research Digest··2 min read
The authors present the first demographically grounded safety evaluation of real-world health conversations with ChatGPT.
§