HAIFA, Israel —From AI and social media to Wikipedia and legacy newsrooms, the final day of “Contemporary Antisemitism 2026” turned to a new frontier in the fight against antisemitism: the digital systems that increasingly shape what people see, believe and what machines learn.

Across multiple sessions dedicated to AI, law, education and quantitative research, speakers described technology as neither inherently dangerous nor inherently beneficial, but rather as a force multiplier for both antisemitic narratives and efforts to analyze, expose and counter them.

‘Artificial intelligence accelerates antisemitism’

Artificial intelligence has not created antisemitism, researchers argued, but it has transformed the speed and scale at which anti-Jewish narratives can spread.

In a panel discussion, Dr. Lev Topor, head of the Cybersecurity Program at the Academic College of Tel Aviv-Yaffo, described artificial intelligence as a new infrastructure through which longstanding prejudices can operate. He warned that AI-generated hate—including synthetic text, manipulated images and deepfake content—has created what he called an “epistemic crisis,” in which audiences increasingly struggle to distinguish authentic material from fabricated content.

“The algorithm isn’t evil, the numbers aren’t evil, the LLM isn’t evil … It is biased,” Topor said. “But evil comes from people.”

To address the challenge, Topor called for practical countermeasures, including AI-powered watermarking, bot detection and regulatory frameworks such as the European Union’s Digital Services Act, which seeks to increase accountability for online platforms.

Researchers warned that existing moderation systems remain poorly equipped to address AI-generated hate at scale. At the same time, they expressed optimism that the same technology creating new challenges could also provide new tools for those combating antisemitism.

AI gives researchers new tools

Haran Shani-Narkiss, founder and CEO of Innohives AI Solutions, argued that AI should not primarily be viewed as a threat. He left academic neuroscience after Oct. 7 to apply his background in data science to media analysis.

“My people need me,” he recalled telling colleagues at University College London. “I just knew that I needed to use my skills and my academic qualifications and everything I gained so far to try and do something.”

His company now uses AI-powered analytical pipelines to uncover patterns across enormous datasets that would previously have been impossible to study manually.

One project examined approximately 1,500 BBC articles published after Oct. 7, identifying what he described as a measurable “sympathy ratio” that shifted dramatically between the broadcaster’s English- and Arabic-language coverage. According to his analysis, the English-language service favored the Israeli narrative by roughly three-to-one, while the Arabic-language service favored the Palestinian narrative by roughly six-to-one.

BBC News Arabic is the second-largest news outlet in the world, behind only Al Jazeera, which has a long record of sharing falsehoods and conspiracy theories about Jews and Israel.

Rather than viewing AI primarily as a danger, Shani-Narkiss argued that the technology gives researchers an unprecedented ability to measure media narratives at scale.

“It’s like having an army of figures working for us,” he said, adding that the analytical power helps address what he described as the Jewish community’s historical numerical disadvantage online.

“I’m putting this back to the question: is it more risk or more opportunity?” he said. “AI is more of an opportunity. It completely deletes the quantity disadvantage that we as the Jewish people have.”

What information is feeding AI?

Beyond the content generated by AI, researchers turned to a deeper question: What information is shaping large language models themselves? As systems such as ChatGPT, Claude and Grok increasingly rely on existing sources of information, researchers warned that legacy media organizations and digital knowledge platforms may play an outsized role in shaping how future audiences—and machines—understand events.

Lilach Sigan, a columnist and researcher at Bar-Ilan University, examined the role of The New York Times as an influential source of information for AI systems. The newspaper’s reputation as a “newspaper of record” means its coverage carries consequences far beyond its own readership.

“It’s a very reliable prime source” for AI systems, she said, calling that influence “a little frightening” given how the outlet covered Israel and Hamas, particularly in the early stages of the war.

Her analysis of 15 months of the newspaper’s daily headline newsletter examined how Israelis and Palestinians were represented in coverage. She argued that positive coverage of Palestinians outpaced positive coverage of Israelis by roughly two-to-one during the first month of the war, while negative coverage of Israel continued to outpace negative coverage of Hamas in later months.

She argued that much of this is driven by the pursuit of short-term revenue through attention-grabbing headlines, but that sensational framing may become embedded in AI systems for years to come.

“Headlines are a very powerful tool in this day and age to frame a story,” she said.

“Negative emotions promote clicks, and the headlines try to bring in more clicks and more revenue,” she said, implying that by “arousing negative emotions,” The New York Times headlines may be interpreted by AI systems as neutral facts.

“Can the New York Times be a reliable source for AI engines?” she asked. “According to my study, the answer is pretty clear that it’s not.”

Researchers also examined another major source shaping digital knowledge: Wikipedia. Shlomit Lir, a researcher at the University of Haifa and Bar-Ilan University, argued that the online encyclopedia’s influence extends far beyond its own platform. The site receives billions of visits annually and is directly used as a source for training artificial intelligence systems.

“Whatever goes into Wikipedia spreads outside Wikipedia,” she said, calling it “a mechanism that shapes public knowledge and perception.”

As part of her research, she interviewed 19 Jewish editors who described witnessing a pattern of contested edits, reversals and informal gatekeeping on Israel-related pages that she termed “knowledge poisoning”: the presentation of a particular narrative “as if it is truth.”

The conflicts, she said, have also taken a personal toll. One longtime editor told her that after prolonged disputes over Israel-related pages, she was hospitalized.

“This really damaged my health, so I stopped,” one editor told Lir. “I won’t touch Wikipedia anymore.”

Lir warned that when Jewish editors withdraw from contested pages, it can further strengthen the influence of other groups shaping what millions of readers—and AI systems—treat as factual information.

When online hate enters the real world

Researchers also warned that the consequences of digital antisemitism are no longer confined to online spaces.

Nilaya Magidish, a research analyst and policy lead at CyberWell, told the AI track the previous day that AI-generated antisemitic content is increasingly moving from the systems that create it onto social media platforms and into public life.

The nonprofit organization combats online antisemitism and uses AI and open-source intelligence to track antisemitic communications.

“The risk here really lies when AI-generated content migrates onto social media platforms,” she said.

She cited a dataset containing more than 30 million views across just 307 posts, demonstrating the speed at which synthetic content can spread. In one case, she presented an AI-generated song calling for violence against Jews that was created following the June 2025 Iran war. The content spread across platforms before its lyrics were later used to harass an Israeli couple traveling in Vietnam.

For researchers, the example demonstrated how AI-generated material can move from digital spaces into real-world consequences. While technology alone cannot solve a social or ideological problem, they argued, the same tools being used to spread antisemitism can also be used to combat it.

The final keynote

The conference concluded with a keynote address by French philosopher, author and public intellectual Bernard-Henri Lévy, whose work has focused on totalitarianism, antisemitism and Jewish identity. Speaking ahead of the Hebrew-language launch of his 2024 book Israel Alone, Lévy argued that while antisemitism continually changes its outward form, its internal logic remains the same.

“The question is not to know if anti-Zionism is or is not a form of antisemitism,” he told the audience. “The real question is: Can antisemitism be something else today than anti-Zionism?”

Drawing on ideas he said he first published five decades ago, Lévy described antisemitism as a “virus” that continually mutates, adapting to new political and cultural environments.

“The only way today to re-attract that old hatred, the oldest hatred … is the way of anti-Israelism.”

His remarks echoed many of the conference’s central themes: that narratives—whether generated by AI or embedded in newspaper headlines and digital platforms—are increasingly shaping how the world understands Israel, the Jewish people and contemporary antisemitism.

“Contemporary Antisemitism 2026,” held in Haifa, brought together more than 220 speakers and more than 500 participants from more than 20 countries. The gathering also marked the launch of the Contemporary Antisemitism Studies Association (CASA), a new academic organization dedicated to advancing interdisciplinary research on contemporary antisemitism.