Racist videos that appear to be created with Google’s AI video generation tool Veo 3 have raked in millions of views across TikTok, according to findings from the nonprofit media watchdog Media Matters. The AI-generated videos uncovered by the organization are filled with racist tropes, many of which target Black people.

Media Matters determined that the videos — one of which received 14.2 million views — were generated with Google Veo 3 because of the “Veo” watermark in the corner of the clips. Some users also listed hashtags, captions, or usernames pertaining to Veo 3 or AI, according to Media Matters. Each clip found by the organization lasted just eight seconds, or was “composed of multiple clips each lasting no more than 8 seconds,” which tracks with Veo 3’s eight-second limit.

Veo 3, launched by Google in May, allows users to generate AI video clips and audio with just a text prompt. On its website, Google says it will “block harmful requests and results.” TikTok’s rules similarly state that “hate speech and hateful behavior has no place on TikTok,” and that the platform “will not recommend content that contains negative stereotypes about a person or group with a protected attribute.”

“We proactively enforce robust rules against hateful speech and behavior and have removed the accounts we identified in the report, many of which were already banned prior to the report publishing,” TikTok spokesperson Ariane de Selliers said in a statement to The Verge.

The Verge found some of the videos highlighted by Media Matters were also on YouTube but with fewer views, while Wired also discovered similar, racist AI-generated videos on Instagram. Other videos found by Media Matters contain antisemitic content and racist depictions of immigrants and Asian people.

Google didn’t immediately respond to The Verge’s request for comment.

Update, July 2nd: Added a statement from TikTok.