Google uses YouTube videos to train AI—but creators don't know
What's the story
Google has admitted that it uses a vast collection of YouTube videos to train its artificial intelligence (AI) models, including Gemini and Veo 3.
The confirmation came after CNBC learned about the practice from a company spokesperson.
However, the tech giant clarified that a subset of its video library is used for training purposes and that it respects agreements with content creators and media companies.
AI training
YouTube's official statement on the matter
A spokesperson for YouTube said, "We've always used YouTube content to make our products better, and this hasn't changed with the advent of AI."
They added that the platform acknowledges the need for safeguards and has invested in strong protections to allow creators to guard their image and likeness in this new era.
However, experts warn that such practices could lead to an intellectual property crisis for creators and media companies.
Training scale
YouTube provides massive training data for Google AI
YouTube has not disclosed how many or which of the 20 billion videos on its platform are used for AI training.
However, experts say that even if only 1% of this massive catalog is used, it would be equivalent to 2.3 billion minutes of content.
This is more than 40 times the amount of data other AI models are trained on.
Lack of awareness
Many creators unaware about this practice
Despite YouTube's claims of transparency, many creators and media organizations are unaware that their content is being used to train Google's AI models.
This lack of awareness was highlighted by CNBC, which spoke to several leading creators and IP professionals who were not informed about this practice.
The revelation comes after Google unveiled Veo 3, one of the most advanced AI video generators on the market.
AI concerns
Creators fear they may be replaced by AI
The use of YouTube videos to train AI models has raised concerns among creators. They fear they may be unknowingly contributing to a system that could one day compete with or replace them.
"It's plausible that they're taking data from a lot of creators who have spent a lot of time and energy into these videos," said Luke Arrigoni, CEO of Loti, a company that protects digital identity for creators.
Terms agreement
YouTube's terms of service grant broad license for content use
YouTube's terms of service state that by uploading content, users grant the platform a broad license to use their videos.
"By providing Content to the Service, you grant to YouTube a worldwide, non-exclusive, royalty-free, sublicensable and transferable license to use that Content," the terms read.
Dan Neely, CEO of Vermillio which helps protect likeness from misuse and secure licensing of authorized content, said many creators are unaware their videos are being used for this purpose.
Copyright issues
Training data could lead to competition with original content creators
Even if Veo 3's final output doesn't directly duplicate existing work, the generated content powers commercial tools that could compete with creators who made the training data possible.
This happens without credit, consent, or compensation.
YouTube has partnered with Creative Artists Agency to give top talent access to identify and manage AI-generated content featuring their likeness.