Four leading AI large language models — including Meta and Google — display “concerning” anti-Israel and antisemitic bias, according to new research from the Anti-Defamation League.