Google is using billions of YouTube videos to train its AI models. As reported in a CNBC article, over 20 billion videos have been uploaded to YouTube, but many creators are unaware that their content is being used. This news is essential because Google uses YouTubers’ data to train personal models, and they cannot opt out of it. Google is exploiting YouTubers’ creativity to train its AI models Gemini and Veo3.
In their defence, a YouTube spokesperson said, “We’ve always used YouTube content to improve our products, and this hasn’t changed with the advent of AI.”
Nevertheless, let’s discuss the whole story and the future implications.
How Google uses YouTubers’ Data to train its AI
Google acquired YouTube in 2006 and has used its data for its products since. The notable shift was when Google Integrated AdSense with YouTube and search algorithms in 2007.
Google has claimed that it uses only a subset of videos to train its AI models. However, many experts believe this violates intellectual property and copyright content.
Let’s look at the other side of the coin. Creators have willingly given the right. The YouTube content license protects users against copyright infringement. However, YouTube has obtained consent from users to use data for its own purposes. Take a look at their terms of service.
Creators have given YouTube access to use their content; now, they can’t opt out of it.
Why Google Uses YouTubers’ Data?
Google’s DeepMind text-to-video model Veo3 launched on May 20, 2025. Recently, the model started developing realistic videos resembling Pixar Studio animation quality. How Veo3 has taken such a giant leap is simple: the highly creative data from YouTube.
If we talk statistically, only 1% of the data from the YouTube library amounts to 2.3 billion minutes of content. Google uses this data, which is 40% more training data than other competitor models.
Other companies, such as Open AI, Salesforce, Anthropic, and Nvidia, have also scraped data from the YouTube Video library. Nevertheless, they were keen on saying “used under copyright law.” Google has now stopped the third-party extraction of YouTube data; however, they haven’t said anything about their own use.
Skeptics also have mixed reviews. Some creators think using their data is unnecessary, and some think creating high-quality content is helpful. Text-to-video creation has almost reached the level of Pixar and DreamWorks. By promoting, people can make high-quality videos.
Conclusion
It has been revealed that Google uses YouTube video library data to train its AI models, Gemini and Veo3. The data amounts to 40% more training data than their competitors, giving them the edge in the AI race. The text-to-video AI models depend on video data; YouTube is vast enough to get that data. Companies like Open AI and Nvidia have already scraped billions of video data from YouTube.
To conclude the news, AI models like Veo3 are drastically changing the text of video generation. In the process, creators’ copyrighted data have been compromised. Although they have agreed to YouTube’s terms of service, it doesn’t make Google the owner of their creative work.