What happens when the world’s most powerful artificial intelligence systems are trained on centuries of human prejudice? A newly published study suggests we are beginning to find out.

Published in American Psychologist , the study, From Myth to Model: Representation of “The Jew” in Generative AI, found that leading large language models consistently associated Jewish identities with a cluster of traits that have long been central to antisemitic narratives. Jewish characters generated by AI were viewed as more competent, privileged, dominant, hierarchical, obsessive, and emotionally controlled, while simultaneously being seen as less warm, less likable, and collectivistic. When researchers translated these profiles into fictional archetypes, the resulting characters closely resembled familiar antisemitic tropes of the manipulative, powerful, morally ambiguous puppet master.

The findings are troubling, but they are hardly surprising. Artificial intelligence systems are trained on vast quantities of human-generated content. They learn from books, articles, websites, social media posts, forums, comment sections, and countless other sources that collectively represent humanity’s knowledge, and humanity’s prejudices. If antisemitism exists throughout society, it will inevitably be reflected in the data that AI consumes.

In that sense, AI is functioning as a mirror, revealing how deeply antisemitic assumptions remain embedded in our culture. The danger is that these assumptions, through AI, will be amplified, legitimized, and distributed at a scale never before possible.

The lesson of the From Myth to Model: Representation of “The Jew”........

© The Times of Israel (Blogs)