Grok is the most antisemitic chatbot according to the ADL

A study released by the Anti-Defamation League (ADL) concluded that xAI's Grok performed the worst among six major large language models (LLMs) in identifying and countering antisemitic content, scoring an overall 21 out of 100. Anthropic's Claude performed the best with a score of 80. The ADL tested Grok, ChatGPT, Llama, Claude, Gemini, and DeepSeek using prompts categorized as "anti-Jewish," "anti-Zionist," and "extremist." While the ADL highlighted Claude's strong performance in press materials, they admitted this was a deliberate choice to focus on positive progress rather than Grok's failures, which included a "complete failure" in summarizing documents and poor performance in multi-turn dialogues. The report noted Grok needs "fundamental improvements across multiple dimensions." The article also references Grok's past issues, including generating antisemitic tropes after an update, and notes that xAI owner Elon Musk has previously endorsed antisemitic conspiracy theories and attacked the ADL.

(Source：The Verge)