Major AI models reproduce centuries-old antisemitic stereotypes, Israeli study finds
Summarized and contextualized by DistantNews.
At a glance
- A new Israeli study found that leading AI models like ChatGPT-4 Turbo reproduce centuries-old antisemitic stereotypes.
- Researchers prompted AI to create biographies for Jewish and non-Jewish names, finding Jewish characters were consistently rated as more competent but less likable and warm.
- The study suggests historical antisemitic tropes, portraying Jews as agents of social disruption, may now be encoded in AI systems.
Leading artificial intelligence models, including ChatGPT-4 Turbo, consistently reproduce centuries-old antisemitic stereotypes, according to a new study by Israeli academics Michael Gilead and Gal Gutman. Their research, titled 'From Myth to Model: Representation of โThe Jewโ in Generative AI,' employed a novel method to identify implicit biases by forming chains of associations within the AI.
The researchers instructed the AI to generate names for Jewish and non-Jewish Americans and then write short biographies for each. When these biographies were evaluated, characters associated with Jewish names were consistently rated higher on competence-related traits like intelligence and assertiveness, but lower on warmth-related traits such as likability and friendliness. This pattern, described as the "high-competence, low-warmth quadrant," was also observed in other AI models like DeepSeek-V3 and Mistral.
From Myth to Model: Representation of โThe Jewโ in Generative AI
The study highlights a concerning persistence of historical antisemitic discourse, which often portrayed Jews as agents of social disruption undermining traditional order. The researchers warn that this ancient prejudice may now be encoded in modern AI systems. They predict that rising anti-modernization sentiments could coincide with an increase in antisemitic discourse, suggesting that the historical association between Jews and the "ailments of modern subjectivity" is not confined to the past but is actively being perpetuated by current technology.
Our analysis reveals how an ancient prejudice persist in modern technolog
Originally published by Jerusalem Post. Summarized and contextualized by our editorial team with added local perspective. Read our editorial standards.