AI models more lax with antisemitism when engaging in Persian, new ADL study finds...

A new study from the Anti-Defamation League highlights that major AI chatbots are less effective at rejecting antisemitic content in Persian compared to English. This inconsistency raises concerns, particularly as millions use these technologies to navigate complex geopolitical issues, including the U.S.-Israel relationship with Iran.

The report underscores the potential dangers of these chatbots perpetuating antisemitism, especially in regions where Persian is spoken. The findings suggest a need for enhanced algorithms to address these safety flaws and improve the handling of sensitive inquiries.