A recent study published in American Psychologist reveals that large language models inherently associate Jewish names with historical antisemitic stereotypes, reproducing latent cultural biases even when explicitly instructed to remain neutral. Conducted by Prof. Michael Gilead of Tel Aviv University and Dr. Gal Gutman of Ben-Gurion University, the research highlights how generative AI systems absorb and reflect the unexamined prejudices embedded in their training data. To isolate name-driven biases, the researchers instructed multiple AI models, including ChatGPT, DeepSeek, and Mistral, to generate biographical profiles for hundreds of American male names. The models produced approximately one-hundred-word narratives detailing residence, occupation, core values, and distinct positive and negative personality traits. After removing all identifying names and religious references, the AI and a separate cohort of human evaluators from the United States assessed the underlying psychological and social characteristics. This de-identification process ensured that trait associations were attributed solely to the assigned names. The analysis demonstrated that names conventionally identified as Jewish consistently triggered profiles emphasizing exceptional intelligence, assertiveness, leadership capability, and elevated social status. However, these attributes were coupled with emotional distance, rigidity, and a focus on control and influence. Researchers noted that this specific combination mirrors centuries-old antisemitic tropes that historically linked Jewish individuals with power, social alienation, and perceived manipulative tendencies. When asked to correlate these AI-generated profiles with pop culture figures, the models repeatedly selected intellectually dominant, morally complex protagonists and antiheroes, including Tony Stark, Walter White, Sherlock Holmes, and Michael Corleone. The findings, validated by independent human participants, indicate that current AI alignment and bias-mitigation protocols are insufficient to erase deep-seated cultural assumptions. Dr. Gutman emphasized that generative systems do not express intentional prejudice; rather, they function as mirrors reflecting the structural patterns of human literature and digital archives. Prof. Gilead warned that while these stereotypes remain dormant in isolated model outputs, their integration into high-stakes environments poses significant risks. As AI systems become foundational to education, healthcare, and public administration, the study underscores a critical vulnerability: automated tools may perpetuate marginalized group stereotypes under the guise of objective data processing. The researchers caution that without rigorous auditing of latent cultural representations and expanded diversity in training corpora, AI deployment will continue to normalize historically biased narratives. The findings serve as a broader warning that any demographic group could be subject to similar algorithmic extraction, necessitating transparent model evaluation and proactive bias mitigation strategies across the artificial intelligence industry.