The dispute centers on Lyria 3, Google’s music-generation model that creates instrumental and vocal tracks from text prompts and powers music features across YouTube and Google’s AI products. While Google has publicly acknowledged using YouTube content to train models such as Gemini and Veo, the company has not confirmed whether YouTube uploads were used to train Lyria. Asked directly about Lyria’s training data, Google declined to comment.
That distinction could become important as the case progresses. Discovery has not yet begun, meaning the plaintiffs have not had an opportunity to examine what data was actually used to train the model. Google’s motion instead focuses on whether the lawsuit can proceed at all under the terms creators agreed to when uploading content.
The company’s legal argument hinges on language in YouTube’s Terms of Service that gives Google rights to “reproduce, distribute, prepare derivative works” from uploaded material. Google is effectively arguing that AI training falls within those permissions. If a court accepts that interpretation, the decision could have implications well beyond YouTube, since many online platforms rely on similarly broad licensing provisions.
Public statements from Google and YouTube have already established that user-uploaded content plays a role in training some of the company’s AI systems. In 2024, YouTube CEO Neal Mohan said that “some portion” of YouTube videos may be used to train models such as Gemini. Later that year, Google said content uploaded to YouTube helps improve experiences across YouTube and Google, including through machine learning and AI applications.
The musicians are expected to argue that those platform licenses were never intended to cover generative AI training. Their position will likely be that language around derivative works and content processing was designed for operational functions such as distribution, playback, or formatting—not for building AI systems capable of generating new music.
The outcome could influence how courts view one of the central questions facing the AI industry: whether platform terms signed years before the generative AI boom are broad enough to authorize model training on user-created content. A ruling in Google’s favor would strengthen the position of companies that own massive content platforms. A ruling against the company could increase pressure on AI developers to negotiate licensing agreements directly with creators and rightsholders.
For now, Google’s strategy is focused on the contractual issue rather than the underlying training data. By seeking dismissal before discovery begins, the company is attempting to resolve the case on the scope of YouTube’s user agreement rather than on whether specific songs were included in Lyria’s development. The court’s response may help determine whether similar challenges against AI training practices gain momentum across the creative industries.
This analysis is based on reporting from AI Chat Daily.
Image courtesy of https://completemusicupdate.com/.
This article was generated with AI assistance and reviewed for accuracy and quality.
