Viral racist and antisemitic TikTok videos appear to have been produced using Google’s latest AI video generator Veo 3, according to Media Matters. The nonprofit research organization found that several of these offensive clips had received hundreds of thousands or even millions of views.
One TikTok video, named “Average Waffle House in Atlanta,” illustrated a restaurant crowded with monkeys throwing watermelons and carrying buckets of fried chicken. It had accumulated over 622,000 views when Media Matters took a screenshot.
Some viewers endorsed the video’s racist stereotypes, with one commenter saying, “all their mannerisms…to the T…”
Another TikTok, posted in mid-June with no fewer than 835,000 views, featured the prompt, “i asked ai: ‘average spirit airlines experience,” and showed monkeys scrambling all over a plane.
Media Matters pointed out that the identified videos were a maximum of eight seconds long, coinciding with the length of Veo 3’s publicly accessible text-to-video clips. The videos were marked “Veo” in the corner or utilized hashtags, captions, or usernames related to Veo 3 or AI, and exhibited errors, distortions, and nonsensical text characteristic of AI-produced videos.
Media Matters provided a compilation of the clips on its YouTube channel.
Mashable contacted both TikTok and Google for remarks on the Media Matters report but had not received a reply at the time of publication.
When Veo 3 debuted in late May, Mashable tech editor Timothy Beck Werth described its realism as both “impressive” and “scary.” Google informed Werth that Veo 3’s measures against misinformation include digital watermarks and adherence to AI safety protocols.
The AI-generated videos identified by Media Matters featured anti-Black stereotypes about criminality, dietary preferences, and absent fathers. Some presented police interactions with Black individuals, including one where a white officer shoots a “Black one” from his vehicle, which had surpassed 14 million views.
The clips also exhibited racist imagery against Asian and South Asian individuals and antisemitic stereotypes, such as Jewish men pursuing a gold coin.
One clip, viewed a million times, depicted a gaunt man in front of a crematorium, vlogging at a Nazi concentration camp, stating, “Well everyone is having a great time here.” It remains uncertain if this minute-long video was created with Veo 3.
Another category of AI-generated video appeared to center on violence against immigrants and those protesting on their behalf.
The videos seemingly breach hate speech regulations enforced by TikTok and Google.
TikTok forbids hate speech and hateful conduct, which encompasses attacking, threatening, dehumanizing, or degrading an individual or group based on their protected characteristics.
Google’s generative AI policy prohibits users from producing or disseminating content that promotes hatred or hate speech, harassment and abuse of others, and violence or the provocation of violence.