AI-generated videos found on TikTok, Meta, YouTube, and X mimicked the popular animation style of Disney/Pixar films to present Holocaust denial and anti-Semitic tropes to children. The image of the second child from the left is from a trailer depicting Jewish children in concentration camps during World War II. Photo: Collage of screenshots from the CyberWell report.
A groundbreaking report released this week reveals that online users have begun abusing AI video generators, weaponizing Disney/Pixar-style nostalgia and packaging toxic hatred in candy-coated shells to appeal to adolescents.
CyberWell, a Tel Aviv-based nonprofit focused on monitoring anti-Semitism on social media, announced on Sunday the results of a study that tracked 307 cases of AI-generated content targeting Jews on social media between January 2025 and February 2026. The group found that the images and videos were viewed 30 million times and had more than 2.8 million user interactions, such as likes and re-shares. They looked at animations created with OpenAI’s Sora, Google’s Veo, X’s Grok, and Suno.
TikTok accounted for the largest portion of the content, around 36%, but the popular video-sharing service also had the highest level of enforcement, at over 88%. Instagram had the highest engagement rate, at around 65%, with anti-Semitic posts totaling nearly 25%.
Metaplatform’s removal rate was 67 percent, significantly higher than Alphabet’s YouTube (28 percent) and billionaire Elon Musk’s X Platform (20 percent). Musk recently incorporated X, xAI, and its Grok chatbot into his rocket company SpaceX ahead of an IPO scheduled for June.
CyberWell discovered three main narratives throughout the video. 33.2 percent portrayed Jews as greedy or obsessed with money, 21.5 percent related to the Holocaust, and 21.2 percent presented violent rhetoric against Jews inspired by a specific event, in this case the June 2025 Israel-Iran conflict and the viral video “Boom, Boom, Tel Aviv.” The video features lyrics such as “Boom, boom, Tel Aviv. This is what you get for all your misdeeds.” […] This is what you brought on yourself, it’s time to shed some blood. […] Humanity has never expected good behavior from you Jews. ”
Researchers call mid-2025 a tipping point in the rise of AI-driven anti-Semitic videos, saying 98.4 percent of identified content will come from that point.
“There is a recurring pattern of users packaging AI-generated anti-Semitic content in formats designed to appeal to younger audiences,” the report said. “The most common examples include fabricated Disney/Pixar-style movie trailers and game-related audio clips that promote Holocaust-related mockery, anti-Semitic conspiracy theories, and hate speech targeting Jews.”
One way users try to circumvent moderation is by tagging such videos with claims of “satire” or “black humor.” Some people use the term “Coast” instead of “Holocaust.”
One of the examples presented is a fabricated trailer for the movie Coast, which was created in the Pixar style using Sora. “Set in a concentration camp, this trailer portrays Adolf Hitler in a light-hearted manner as it follows a group of Jewish boy prisoners as they attempt a dramatic escape. By presenting the Holocaust in a playful animated format, the video turns atrocities into entertainment and diminishes the gravity of Jewish suffering.”
AI videos also tap into children’s love of video games.
The TikTok video he created with Sora titled “CAUST COMMANDER” received 66,500 views, 4,623 likes, and 3,619 reposts. “The post depicts Adolf Hitler murdering those around him in a playful, stylized manner. The video downplays the Holocaust and the mechanisms used to exterminate the Jewish people by presenting it in a gamified, commercialized format, including fake products such as Zyklon B gas, themed costumes, and promotions such as ‘sparkle on the back,'” the report said.
On March 24, OpenAI announced its decision to shut down Sora following months of reports of anti-Semitic content spread across the platform. Analysts determined that the decision was motivated by the need to free up computational power for training new models so that OpenAI could remain competitive as Anthropic’s Claude soared in popularity among programmers and Alphabet’s Gemini alienated users.
The report highlights that anti-Semitism is deeply embedded in AI systems.
“Many of the websites used in the AI training dataset serve as active hubs of anti-Semitic discourse, raising concerns about their inclusion in model development,” the report states. “For example, Reddit ranks as one of the most cited domains across major AI systems, yet analysis of ChatGPT output shows that Wikipedia alone contributes to about half of the responses generated. AI companies’ reliance on these websites highlights the risk that anti-Semitic discourse circulating online can be embedded in model inputs and later disseminated at scale.”
Tal-Or Cohen Montemayor, CEO and founder of CyberWell, warned that AI has accelerated both the speed and intensity of online anti-Semitism.
“Artificial intelligence has fundamentally changed the scale and speed at which anti-Semitism is produced and distributed online,” said Cohen Montemayor. “Generative AI allows bad actors to industrialize hate and create high-impact content that can reach millions of people. Enforcement often occurs only after hate has already been widely amplified.”
Cohen Montemayor added that CyberWell’s latest report “examines the circulation of anti-Semitic AI-generated content on major platforms and provides important insights into how social media platforms can address the misuse of generative AI tools to spread anti-Semitism in the digital world.”
CyberWell found that while the disarming aesthetic created by Pixar targeted children, it was videos that openly glorified violence that drove the highest shares, reaching 33% of content but 41% of engagement.
Cohen Montemayor called on platforms to “move beyond disclosure and invest in systems that identify harmful narratives at scale, including those embedded in audio, video, and coded formats that evade traditional detection.”
Cohen-Montemayor warned that AI is being “weaponized at scale,” explaining that “through continued collaboration with technology companies, policymakers, and specialized partners, we can address complex and rapidly evolving forms of online hate by strengthening automated detection, investing in competent and transparent human moderation, auditing training data, and partnering with external expert stakeholders, platforms, and AI developers.”
In one of the most-watched legal battles in the artificial intelligence field, a jury on Monday ruled against Musk in a lawsuit he brought to force OpenAI to fully return to its original nonprofit mission. Jurors decided that Musk filed the lawsuit too late.
The world’s richest man could face tougher legal challenges over his AI business in France, with prosecutors saying they intend to bring criminal charges against the Grok chatbot for promoting Holocaust denial and the generation of child sex abuse images.
