Lawsuit accuses Google of training Lyria AI on YouTube songs without permission

A group of independent musicians has filed a lawsuit against Google, accusing the company of training its music-generation AI on their songs without permission. Google is now asking the court to throw the case out, pointing to YouTube's terms of service as its main line of defense.

According to The Verge, the artists, who all publish their work on YouTube, say their music was used to train Lyria 3, Google's generative model that creates songs from text prompts. None of them, they claim, ever gave permission for that.

Google's response leans on two arguments. The first is that the lawsuit is built on assumptions the plaintiffs cannot actually prove. The second, and far more consequential, is that anyone who uploads to YouTube has already signed off on the kind of use being challenged. The platform's terms grant it a broad license to reproduce, distribute, and build on uploaded content, and Google argues that even if specific songs were used in training, doing so falls inside that existing license.

This is not the first time Google has acknowledged using YouTube to train AI. In 2024, YouTube CEO Neal Mohan told Bloomberg that parts of the platform's video catalog were going into models like Gemini, and the company has since confirmed that the same library helped train its Veo video generator. What Google has so far refused to say publicly is whether Lyria was built the same way.

Legal analysts see Google's silence around Lyria as a deliberate move. While the case is still active, any specific admission about how the model was trained could set a precedent the company wants to avoid. The lawsuit also reflects a broader standoff between the creative industry and AI developers, where one central question remains unsettled in court. Whether AI companies can legally train their models on copyrighted work without consent or compensation is still being decided.