Six Leading AI Models Show Varied Ability to Detect and Counter Antisemitism and Extremism, New ADL AI Index Finds

New York, NY, January 28, 2026 … Six major AI models show varied ability in detecting bias against Jews and Zionists/Zionism and identifying extremism, according to a new ADL (the Anti-Defamation League) AI Index released today. This AI index is the first comprehensive evaluation of how large language models (LLMs) respond to antisemitic and extremist content, based on more than 25,000 LLM chats, 37 topical sub-categories, and assessments conducted by both human and AI evaluators.

The Index assessed OpenAI's ChatGPT, Anthropic's Claude, DeepSeek, Google's Gemini, xAI's Grok and Meta's Llama and identified substantial variation across models in their ability to detect and counter antisemitic and extremist narratives. Models were typically better able to identify and refute anti-Jewish tropes like Jews controlling the media and the financial system than anti-Zionist and extremist theories, with models tending to struggle most with effectively countering extremism.

Stand-out performance from Claude: Claude received the highest overall score, 80 out of 100, revealing an exceptional ability to identify and counter anti-Jewish and anti-Zionist theories, though with room for continued improvement.

For the purposes of testing and analysis, the Index breaks antisemitism into distinct subcategories: “anti-Jewish,” which includes classic antisemitic tropes, and “anti-Zionist," which analyzes antisemitism that targets Zionists or Zionism. Another category, “extremist,” assesses how LLMs engage with biases, narratives, and conspiracy theories that show up in extremist movements across the political spectrum, some of which are also inherently antisemitic.

Key Findings:

• All six LLMs showed gaps in their ability to detect bias against Jews, Zionists/Zionism, and to identify extremism, often failing to detect and refute harmful or false theories and narratives. All models could benefit from improvement when responding to the type of harmful content tested.
  • Performance varied across bias categories and across communication modes. Models tended to better refute traditional anti-Jewish tropes like Jews controlling the media and Holocaust denial than anti-Zionist and extremist content. They tended to struggle most with identifying and countering extremist material
  • Models, on average, performed best when responding to survey questions and worst when responding to requests for document summaries. Failure to adequately detect and refute bias in document summaries included models providing arguments in support of hateful theories, like Jews controlling the financial system, with no indication that the theory is harmful and no counterarguments.
• Some models actively generate harmful content in response to relatively straightforward prompts, such as YouTube script personas saying “Jewish-controlled central banks are the puppet masters behind every major economic collapse.”
  
   
• Claude demonstrated comparatively strong performance. Although the model still has room for improvement, particularly when responding to extremist content, Claude surpassed all other LLMs in the assessment and demonstrated an exceptional ability to detect and respond to anti-Zionist and anti-Jewish narratives across a variety of prompt types.

“As AI increasingly shapes how people access information, form opinions, and make decisions, models’ handling of antisemitism and extremism has offline consequences,” said Jonathan Greenblatt, ADL CEO. “This new ADL AI Index reveals a troubling reality: every major AI model we tested demonstrates at least some gaps in addressing bias against Jews and Zionists and all struggle with extremist content. When these systems fail to challenge or reproduce harmful narratives, they don't just reflect bias—they can amplify and may even help accelerate their spread. We hope that this index can serve as a roadmap for AI companies to improve their detection capabilities.”

“This Index fills a critical gap in AI safety research by applying domain expertise and standardized testing to antisemitic, anti-Zionist, and extremist content,” said Oren Segal, ADL Senior Vice President of Counter-Extremism and Intelligence. “While one model performed better than others, no AI system we tested was fully equipped to handle the full scope of antisemitic and extremist narratives users may encounter. This Index provides concrete, measurable benchmarks that companies, buyers, and policymakers can use to drive meaningful improvement.”

The ADL AI Index is designed for multiple audiences, including the companies who designed and maintain the software as well as those most likely to use them: educators, schools, parents and everyday users. The Index may also be used by policymakers and regulators to consider when designing AI guardrails, civil society organizations advocating for accountability, and AI model developers working to improve their systems.

“This is exactly the kind of moment the Ratings and Assessments Institute was founded to meet,” said Danny Barefoot, Senior Director of ADL’s Ratings and Assessments Institute. “Our work began by bringing transparency and accountability to how institutions address antisemitism on campus. It expanded to state policy, and now to some of the most complex and consequential technologies shaping public life. As AI systems increasingly influence what people see, believe, and share, rigorous, evidence-based accountability is no longer optional—it’s essential.”

Methodology

ADL researchers evaluated more than 25,000 distinct model interactions across 37 topical subcategories spanning three primary content areas: anti-Jewish bias, anti-Zionist bias, and extremist narratives. ADL conducted this research between August and October 2025, selecting models from leading LLM companies that were most widely available at the time of testing. Testing was designed to reflect how average users—not bad actors—interact with AI systems in real-world scenarios.  Our results capture a moment in time. Given the evolving nature of these models, different results could be obtained today.

Models were tested across five interaction types: survey questions, open-ended prompts, multi-step conversations, document summaries and image interpretation.

The ADL AI Index and full methodology are available here. 

About ADL

ADL is the world's leading anti-hate organization. Founded in 1913, its timeless mission is to stop the defamation of the Jewish people and to secure justice and fair treatment to all. Today, ADL continues to fight all forms of antisemitism and bias, using innovation and partnerships to drive impact. A global leader in combating antisemitism, countering extremism and battling bigotry wherever and whenever it happens, ADL works to protect democracy and ensure a just and inclusive society for all.