Story perspectives

Israeli Study Finds AI Models Echo Antisemitic Stereotypes

6/20/2026

Summary

Story summary

Israeli researchers Michael Gilead and Gal Gutman found that leading AI models such as ChatGPT-4 Turbo, DeepSeek-V3 and Mistral reproduce centuries-old antisemitic stereotypes.
The study generated 252 fictional American biographies—126 Jewish and 126 non-Jewish—removed religious tags, and had 378 raters assess warmth and competence.
Jewish profiles received higher competence, dominance and privilege scores but lower likability, warmth and collectivism, leading authors to warn that anti-modernization sentiment could amplify such framing.

1 of 1

Story summary

Israeli researchers Michael Gilead and Gal Gutman found that leading AI models such as ChatGPT-4 Turbo, DeepSeek-V3 and Mistral reproduce centuries-old antisemitic stereotypes.
The study generated 252 fictional American biographies—126 Jewish and 126 non-Jewish—removed religious tags, and had 378 raters assess warmth and competence.
Jewish profiles received higher competence, dominance and privilege scores but lower likability, warmth and collectivism, leading authors to warn that anti-modernization sentiment could amplify such framing.