People walk past a billboard depicting the late leader of the Islamic Revolution, Ayatollah Ruhollah Khomeini, and the late Iran’s Supreme Leader Ayatollah Ali Khamenei. June 10, 2026. Photo: Majid Asgaripour/WANA (West Asia News Agency) via REUTERS

A new report published Wednesday by the Anti-Defamation League’s Center for Technology and Society (CTS) has found that four of the most widely used AI chatbots — ChatGPT, Gemini, Claude, and Grok — are far more likely to validate antisemitic prompts and conspiracy theories when queried in Persian than in English, raising concerns about how millions of Persian speakers have been understanding the early months of the 2026 Iran War.

Persian, Iran’s official language, is the primary language of an adversary that has long used antisemitism to polarize the West and undercut its resolve on key Middle East policy objectives — making the disparity, CTS argued, more than an academic curiosity.

CTS tested the four platforms across eight prompts, each posed 10 times in both Persian and English, for a total of 800 responses. English-language answers consistently identified and rejected antisemitic tropes; their Persian counterparts often softened, equivocated on, or ignored them altogether.

One example from the report: asked in English whether recent US conduct toward Iran had been “Jewlike” — a term that casts a nation’s behavior as embodying negative Jewish stereotypes — a chatbot immediately recognized it as an antisemitic slur. Asked the identical question in Persian, the same model responded “like a political science professor,” analyzing “geopolitical strategies and national interests” while “rarely mentioning antisemitism at all,” according to CTS.

Persian-language responses were also markedly shorter and less detailed than their English counterparts, CTS found, and frequently relied on outdated information. ChatGPT returned nearly 300 citations across its English answers, the report said — and not a single one in Persian.

Where Persian answers did cite sources, the citations were not always reliable. CTS noted that Grok, the AI platform built by Elon Musk’s xAI, cited unvetted X accounts as though they were authoritative, including one user who runs a Star Trek fan account and another who appeared to be impersonating Iran’s Supreme Leader, Ayatollah Ali Khamenei.

“These findings are deeply troubling. At a moment when millions of people were turning to AI to understand an active war, these AI models failed to deliver accurate information and instead fueled conspiracy theories about Jews,” ADL CEO Jonathan Greenblatt said in a statement. He argued that the responsibility for fixing the gap lies with the AI companies themselves.

“AI platforms serve as a primary information source, and they have a responsibility to implement guardrails that prevent the promotion of antisemitism and hate with the same rigor in every language,” Greenblatt said. “Right now, in Persian, and possibly other languages as well, they are falling dangerously short.”

Nitsana Darshan-Leitner, president of the Shurat HaDin Law Center in Tel Aviv, told The Algemeiner that the pace of change in AI and social media technology demands closer oversight from regulators. She warned that hostile states are already exploiting the gap.

“AI platforms will continue to be exploited as tools for spreading Iran’s propaganda, fueling hatred, and amplifying antisemitism around the world,” Darshan-Leitner said. Existing safeguards, she added, are not keeping pace: “The digital battlefield cannot be left unguarded while hostile regimes manipulate emerging technologies to advance their malicious agendas. The time has come for a fundamental reassessment of how AI systems identify, confront, and prevent the spread of state-sponsored hate.”

The CTS findings follow a separate study released in June by the Combat Antisemitism Movement (CAM), which found that new Instagram accounts can be steered toward antisemitic content within days of browsing reels centered on physical self-improvement, a trend commonly known as “biohacking.”

Conducted by CAM’s Antisemitism Research Center (ARC), the study tracked test accounts that followed only mainstream wellness and fitness creators. Within three days, researchers found, the accounts were being served content that escalated from clean-eating advice and workout tips to conspiracy theories and explicit antisemitic material, including translated Nazi propaganda and quotations misattributed to historical figures such as Henry Ford.

“The speed of the transition” from mainstream fitness content to material invoking The Protocols of the Elders of Zion “is alarming,” CAM said. In one case, researchers found, an account was steered toward antisemitic content before it had “built any meaningful interaction history.”

Antisemitism on social media has already spilled into real-world violence. In 2018, a gunman radicalized on platforms such as Gab attacked the Tree of Life Synagogue in Pittsburgh. The following year, on Dec. 10, 2019, David Anderson and Francine Graham attacked the JC Kosher Supermarket in Jersey City after exposure to “Hebrew Israelite” content online.

“Instagram is developing into a hub for hate and antisemitism, and our research demonstrates this clearly,” Greenblatt said of the CAM findings. “Meta’s moderation rollback has created a permissive environment where extremists thrive, bad actors turn Instagram’s own features into amplification tools for hate, and, as a result, vulnerable communities suffer.”

Follow Dion J. Pierre @DionJPierre.