AI chatbots consistently fail to identify, reject antisemitism in Persian

New research tests four major platforms across 800 responses, exposing stark disparities between English and Persian language safety guardrails.