AI platforms display considerable variance in judging hate speech, Annenberg School study finds
The researchers tested seven models — including those developed by OpenAI, DeepSeek, and Google — and found that the models display inconsistencies when identifying and evaluating hate speech.
















