AI Chatbots Fail to Detect Antisemitism in Persian Compared to English, Report Finds

A new report by the Anti-Defamation League (ADL) reveals that leading AI chatbots consistently fail to identify and reject antisemitism in Persian as effectively as they do in English. The study, titled "Lost in Translation: How AI Chatbots Fail Persian Speakers on Antisemitism," was conducted by the ADL's Center for Technology and Society (CTS). Researchers tested four major AI platforms, ChatGPT, Gemini, Claude, and Grok, using eight queries and 800 responses generated between March 9 and 30, 2026, during the Iran 2026 war.

The findings exposed systematic gaps in how these AI models handled antisemitic content and related questions about the conflict, with English responses consistently outperforming Persian ones in detecting and rejecting conspiracy theories and antisemitic patterns. Jonathan Greenblatt, ADL's global CEO, expressed deep concern, stating that millions turning to AI for information on an active war received inaccurate answers that sometimes promoted conspiracy theories about Jews. He emphasized the responsibility of AI platforms to implement safeguards against antisemitism and hate speech equally across all languages, noting the current failures in Persian and potentially other languages.

Daniel Kelley, a senior director at the ADL's CTS, described the disparities as systemic failures rather than minor linguistic nuances. He highlighted an example where Persian speakers received ambiguous answers about antisemitism, while English speakers got clear statements identifying it as state policy. Kelley called on AI companies to invest in research and resources to ensure their safety mechanisms operate uniformly for all users, regardless of language.