Anti-Semitism Flourishes with New Artificial Intelligence Voice Mimicry
The on-line video shows a beautiful, sexy woman promising unconditional love while another depicts Adolf Hitler screaming, in full color, in English, about the existential danger of Jews to Germany and the world.
To the untrained eye, both videos seem real but neither is real. Both are wholly computer generated, two extremes in the burgeoning world of artificial intelligence (AI). The danger isn’t that made-up women will offer realistic unquestioned love. The real menace is that the proliferation of AI Hitler will fuel further anti-Semitism and white power bigots.
It is already happening and becoming more vile.
Musk bought Twitter in 2022, renaming it X. Within hours, the site received a wave of racist slurs and Nazi memes that have continued. Analysts with the Institute for Strategic Dialogue (ISD) have identified hundreds of videos featuring Hitler’s speeches, translated into English using artificial intelligence. The videos, in some cases posted as early as January 2024, have garnered millions of views.
On TikTok, most of the identified videos included 20–30-second audio clips of Hitler’s speeches in English along with background music. The videos often included anti-Semitic, anti-Muslim, anti-immigrant and white supremacist messages.
Hitler posts have been praised on X and TikTok, with comments like “I miss you uncle A,” “He was a hero,” and “Maybe he is NOT the villain.”
In one video, Hitler’s patented AI voice quavers and crescendos as he delivers English-language versions of his most notorious addresses, including his 1939 Reichstag speech predicting the end of Jewish people in Europe. Pro-Trump conspiracy theorist Dominick McGee posted to X an English-language AI audio recreation of the Reichstag speech, which garnered 13,000 retweets, 56,000 likes and more than 10 million views.
McGee’s X account was reinstated last year by Musk after he was suspended for posting a video of child sex abuse.
Among other AI videos, one featured a silhouette of Hitler with the message: “Growing up is realizing who the villain really was.” The video garnered 548,000 views before being removed by TikTok.
Similar short audio clips were available on Instagram, with more than 140 short-format videos featuring AI-translated Hitler speeches.
“Dancing Israelis” is the title of a video posted on X. The credits claim it is a “Mossad/CIA production.” At first, it appears to be an animated, Pixar creation but its content is anything but. It shows a caricatured stereotype of a dancing Jewish man whose boot is knocking down the World Trade Center towers.
“Dancing Israelis” was posted by a user calling himself, “American First patriot,” who has about 220,000 followers. The video had 190,000 views with 8,000 links and 1,500 re-shares. The post was repeated, with 194,000 more views. X took no action against the tweet, according to a report from the non-profit Center for Countering Digital Hate (CCDH).
Many of the most vile images remained on line even though an examination by the ISD found that much of the content appears to violate X’s policies, which states: “We consider hateful imagery to be logos, symbols, or images whose purpose is to promote hostility and malice against others based on their race, religion, disability, sexual orientation, gender identity or ethnicity/national origin. Some examples of hateful imagery include but are not limited to: symbols historically associated with hate groups, e.g., the Nazi swastika.”
ISD said the persistence of pro-Hitler content may be caused by detection failures within moderation protocols, unclear definitions within policies or the misapplication of exceptions social media platforms put in place for content deemed to be “in the public interest.”
“All platforms assessed include policy carve-outs for educational or newsworthy content, yet these definitions are often vague and subjective, with a lack of transparency as to when they have been applied,” ISD found.
“Dancing Israelis” is part of a new wave of AI-generated hate memes that were re-posted on X after they were initially posted on 4chan, an anonymous online messaging board that has long attracted offensive and extremist content. Research by the CCDH found that a campaign by 4chan members to spread “AI Jew memes” after the deadly Oct. 7 Hamas attack against Israel resulted in 43 different images reaching a combined 2.2 million views on X between Oct. 5 and Nov. 16.
The CCDH research showed the most widely shared AI tweet was titled, “Pixar’s Nazi Germany,” and included a montage of four scenes from an imaginary movie with smiling Nazis running concentration camps and leading Jewish children and adults into gas chambers. As of last week, the post had been viewed more than half a million times.
Other AI posts showed U.S. soldiers kneeling before a Jewish man on a throne; Taylor Swift in a Nazi officer’s uniform sliding a Jewish man into an oven; and a Jewish man pulling the strings on a puppet of a Black man.
The CCDH reported that the verified user who tweeted the image of soldiers bowing to a Jewish ruler, included a tweet claiming that Jews seek to enslave the world. The user was not identified other than he ran for U.S. Senate in Utah as a Republican in 2018 and has 86,000 followers on X.
Some of the same AI-generated images also have been posted to other TikTok, Instagram, Reddit, YouTube and Facebook, but the CCDH said X has been a favored platform.
The newer posts are reminiscent of the recent resurrection of Osama Bin Laden’s propaganda justifying the September 11 attacks.
Since the propaganda was made available, platform users began questioning the known facts of World War II and leaning into conspiratorial narratives questioning America’s school curriculum on Adolf Hitler and his role in the Holocaust.
A November 2023 ISD report noted that the legacy of bin Laden was trending in the wake of the Israel-Hamas conflict because of his “Letter to America,” a 2002 screed that justified al-Qaeda’s attacks on the United States. Over the course of November 14–16, references to bin Laden on X jumped more than 4,300 percent, from just over 5,000 to more than 230,000 references. References to “Letter to America,” increased by more than 1,800 percent, from just over 4,800 to 100,000, with 719 million impressions across the platform.
On TikTok, ISD researchers found 41 “Letter to America” videos with more than 6.9 million views (available to minors as well as adults). While TikTok has now blocked “Letter to America” it is still easily accessible under the search term “Bin Laden.” Platforms such as TikTok began reacting to the trend by blocking search terms such as “Letter to America.” ISD researchers, however, found that the content would still appear on TikTok through a simple search for “Osama bin Laden.”
Researchers found the full text of the speech in English and Arabic avoided moderation across X, Facebook and TikTok.
“The case of ‘Letter to America’ signifies a societal failure in many regards, but the platforms are as much to account for its spread as the users,” the ISD report noted.
Content identified by ISD since August included an AI video of Hitler with children, referring to Nazi Germany as a “magical place,” a video compilation of “Sieg Heil” salutes to Hitler where the post’s author asks for users to “Comment ‘sieg heil’ below,” an image of Hitler clutching his heart with the post reading “Hitler seeing more and more people see he was right” and an image of Hitler below the text “the answer to 2024 is 1934.”
On X, posts featuring AI-generated Hitler speeches often included real footage of the speeches. One post from February 2024 included real footage and AI-translated audio of Hitler’s 1935 speech at Krupp munitions factory. It gained 7.9 million views and remains available on the platform.
The post was easily viewed on You Tube, with comments including, “The man is one of a kind in his speeches. He wanted the best for Europe. Heil !”, “My glorious king Adolf has affected many lives very deeply. And I truly believe he never did any wrong. I wish he was still alive to touch me I am blonde and I have blue eyes which i KNOW MY GLORIOUS KIND WOULD ABSOLUTELY ADORE” to “A man like this is only born once every 100 years, a real visionary, Germany rose up like a phoenix from its ashes under this man’s rule.”
The ISD found that in several X and TikTok accounts, users were directed to YouTube channels where longer-form versions of these speeches were available. One of the videos, featuring a 1922 Hitler speech, garnered 1.2 million views and more than 16,000 comments. Analysts also found at least one tutorial video demonstrating how to use AI-powered voice cloning software to recreate Hitler’s voice in English
A study of trends in anti-Semitic postings was released last week by ISD analysts. Among its finding:
· Content glorifying Hitler, questioning facts about Hitler’s intent, or including English audio versions of his speeches, received more than 50 million views across X, YouTube, TikTok and Instagram throughout 2024.
- Content was more easily accessible and had significantly higher reach on X than TikTok, Instagram or YouTube.
- The content’s audio predominantly featured Hitler’s speeches translated into English via AI. On X, the visuals were primarily of Hitler; on TikTok the videos often included imagery of young, western influencers, landscape images and religious symbols.
- Seven AI-generated videos of Hitler speeches on YouTube, posted in 2024, had received 6.9 million views by the end of data collection.
- Since August 13, posts that appeared to glorify or support Hitler, or included Nazi iconography, received more than 24.8 million views across X, TikTok and Instagram.
- The content seemed to be more readily available and attained higher reach on X, with just 11 posts garnering 11.2 million views in a one-week period.
A number of major companies have pulled ads on X because of tweets promoting anti-Semitism. They include IBM, Apple, the Lionsgate film production company, Disney, Paramount, Comcast and Warner Bros.
The AI technology is advancing quickly and become ever more affordable. fast. The latest version from Midjourney costs $30 per month. Companies such as ElevenLabs transform a short vocal sample into a synthetically generated voice through a text-to-speech tool. ElevenLabs software can be free or cost between $5 and $330 per month to use, with higher prices allowing users to generate more audio.
ElevenLabs drew criticism after its tool was used to replicate the voice of the actor Emma Watson falsely reciting passages from Adolf Hitler’s “Mein Kampf.”
Days before a crucial 2023 national election in Slovakia, an audio clip began circulating with a voice that sounded like the country’s Progressive party leader, Michal Šimečka. The clip had Šimečka outlining a scheme to rig the vote, in part by bribing members of the country’s marginalized Roma population. The clip was soon debunked by fact-checkers as likely fakes.
Recently, the actor Tom Hanks warned his social media followers that bad actors used his voice to falsely imitate him hawking dental plans. Over the summer, TikTok accounts used AI narrators to display fake news reports that linked former president Barack
Social media companies find it difficult to moderate AI-generated audio partly because human fact-checkers often have trouble spotting fakes. Previously, voice cloning software churned out robotic, unrealistic voices. But technology now can analyze millions of voices, spot patterns in elemental units of speech and replicate it within seconds.