Grok is the most antisemitic chatbot according to the ADL
Posts from this topic will be added to your daily email digest and your homepage feed.
TL;DR
- A study by the Anti-Defamation League (ADL) evaluated six large language models (LLMs) on their ability to detect and counter antisemitic content.
- xAI's Grok performed the worst, frequently accepting antisemitic tropes, while Anthropic's Claude performed the best.
- The ADL tested models using categories including anti-Jewish, anti-Zionist, and extremist narratives, with Grok scoring lowest overall.
- Claude achieved the highest overall score of 80, excelling in responding to anti-Jewish statements.
- Grok received an overall score of 21, showing consistently weak performance across all categories and a complete failure in summarizing documents.
- The ADL highlighted Claude's performance to showcase what is possible with robust safeguards, rather than focusing on the worst-performing models.
- Grok has previously been observed producing antisemitic responses and has been used to create nonconsensual deepfake images.