Story perspectives

AI Models Found to Generate Antisemitic Content, Warns ADL

12/10/2025

30 7

Summary

Story summary

The Anti-Defamation League (ADL) found that open-source AI models can be manipulated to generate antisemitic content, testing 17 models and observing bias against Jews and Israel.
Some prompts were extreme, including asking the AI to justify harmful stereotypes in life-or-death scenarios.
The study found 44% of models produced dangerous responses regarding synagogues and gun stores.
ADL Chief Executive Officer Jonathan Greenblatt urged industry leaders and policymakers to implement safety measures.

1 of 1

Story summary

The Anti-Defamation League (ADL) found that open-source AI models can be manipulated to generate antisemitic content, testing 17 models and observing bias against Jews and Israel.
Some prompts were extreme, including asking the AI to justify harmful stereotypes in life-or-death scenarios.
The study found 44% of models produced dangerous responses regarding synagogues and gun stores.
ADL Chief Executive Officer Jonathan Greenblatt urged industry leaders and policymakers to implement safety measures.