Videos featuring racist content, possibly created using Google’s AI tool Veo 3, have gained millions of views on TikTok. The videos reportedly include negative stereotypes.

Media Matters identified the use of Veo 3 through watermarks, hashtags, captions, and usernames. One video reportedly reached 14.2 million views.

Google Veo 3, launched in May, allows users to generate short videos and audio from text prompts.

Although Google claims the tool filters harmful content, videos also contained antisemitic content and offensive portrayals of immigrants and Asians.

A TikTok spokesperson confirmed that accounts mentioned in the report have been removed, some even before the report’s release.

Similar videos were also found on YouTube and Instagram, though with fewer views.

Google has not responded to requests for comment.

🔗 Source: The Verge

🧠 Food for thought

1️⃣ AI systems consistently reproduce racial biases despite safeguards

Research shows AI models exhibit significant bias against speakers of African American English (AAE), assigning them lower-prestige jobs and higher conviction rates in hypothetical criminal cases compared to speakers of standardized American English 1.

These biases in AI reflect historical stereotypes, with language models producing systematically negative evaluations of AAE speakers 2.

The current situation with Veo 3-generated racist videos mirrors previous incidents like Microsoft’s Tay chatbot, demonstrating a persistent pattern where AI systems absorb and amplify societal biases despite technical safeguards.

University of Chicago researchers found that even when AI systems generate overtly positive associations for certain groups, they simultaneously produce covert negative stereotypes based on linguistic patterns, highlighting the sophistication of modern algorithmic bias 1.

Current bias mitigation practices, including human feedback training, may inadvertently reinforce rather than eliminate these covert biases, suggesting fundamental challenges in creating truly unbiased AI systems 2.

2️⃣ Content moderation struggles to keep pace with AI-generated hate

Recent Google developments