Multilingual safety guardrails matter because models deployed globally can behave very differently across languages, creating moderation and trust risks for non-English users. According to a new report from the Anti-Defamation League (ADL), the world's most widely used AI chatbots "fail, without exception, to identify and reject antisemitism as effectively in Persian as they do in English." The ADL Center for Technology and Society tested ChatGPT, Gemini, Claude, and Grok across eight prompts, generating 800 responses between March 9-30, 2026, during the 2026 Iran War, and found Persian answers often hedged, softened, or failed to recognize antisemitic content while English answers rejected it outright. ADL CEO Jonathan Greenblatt called the results "deeply troubling." Reporting in The Jerusalem Post and JNS/Cleveland Jewish News echoes ADL's findings.
Yale Professor Warns AI Companions May Harm Social Skills