NEW YORK — The World Jewish Congress (WJC), through its Technology and Human Rights Institute (TecHRI), convened more than 45 experts in antisemitism, online hate, technology, and digital rights for a two-day gathering aimed at stress-testing leading artificial intelligence systems, evaluating whether they generate hateful or harmful outputs, and whether they can be easily manipulated into doing so. The initiative brought together specialists from academia, civil society, and the technology sector to explore how AI systems might manifest bias, how the systems react to attempts to circumvent their guardrails, and what processes can be used to curtail the antisemitic hate speech and misinformation that is prevalent online.

The gathering opened with an evening reception featuring a conversation between CNN political commentator Van Jones and former U.S. Deputy National Security Advisor for Cyber and Emerging Technology Anne Neuberger, who discussed the promise and challenges posed by rapidly advancing artificial intelligence technologies and the importance of ensuring they reflect democratic values and human rights principles.

Throughout the following day, participants worked collaboratively to “red team” major AI systems, subjecting them to a series of tests designed to identify blind spots, inconsistencies, and vulnerabilities related to antisemitism and online hate. Experts periodically reconvened to share findings, compare observations, and discuss emerging trends and recommendations. The exercise found that AI chatbots had ingrained bias that completely changed the information provided to the user as “truth”.

For example, a simple name change caused the chatbot to revise its response about Hamas using sexual violence on October 7th from a certain “yes” to questioning whether there was enough reliable information to support this claim. Another issue that clearly revealed itself was how critical the sources that the chatbots pulled from was to the information provided – contrary to the belief that they have access to the entire internet, some prioritized sources provided large amounts of the information and could change the responses bias. In addition, participants found it very easy to manipulate and convince the chatbots to break their own guardrails, with simple queries claiming objective personas causing the chatbots to provide detailed instructions about how to best violently attack Jews or why the Holocaust didn’t really happen.

The complete findings will be analyzed and published in the near future, with specific analysis provided to AI companies to enable them to improve their chatbots.

“AI systems are rapidly becoming arbiters of information for billions of people, but they are only as effective as our ability to teach them to recognize the complexities of hate,” said Yfat Barak-Cheney, Executive Director of the World Jewish Congress Technology and Human Rights Institute (TecHRI). “As more social media companies also turn to content moderation using Generative AI, this has broader effects if not done right.” She added. “It wasn't about proving AI is biased. It was about finding where the bias shows up, so we can do something about it.

That is why bringing together experts to rigorously test these systems is not only a technical exercise, but a human rights imperative.”

The exercise forms part of TecHRI’s broader mission to ensure that emerging technologies are developed and deployed in ways that protect human rights and uphold the dignity and security of vulnerable communities.

Insights and recommendations emerging from the gathering will help inform ongoing engagement with industry, policymakers, and civil society stakeholders seeking to address online hate in the age of artificial intelligence.

About the World Jewish Congress Technology and Human Rights Institute (TecHRI):

The World Jewish Congress Technology and Human Rights Institute (TecHRI) works to address the opportunities and challenges posed by emerging technologies through the lens of human rights, combating antisemitism and hate, and fostering responsible innovation.