Google tells court YouTube terms give it a “broad license” to train AI music on songs uploaded by indie artists

Google has asked a federal court to dismiss a copyright infringement lawsuit brought by independent artists over Google’s AI music model Lyria 3, arguing that YouTube’s Terms of Service give Google a contractual right to train AI systems on music uploaded to the platform. The dispute centers on whether the creators’ upload activity amounts to permission for AI training, rather than whether the training was lawful under copyright’s typical defenses such as fair use.

According to Google’s filing summarized by Billboard, the company contends that when indie artists upload songs to YouTube, they grant a “broad license” covering the training of artificial intelligence models using that content. Google made that argument in a motion to dismiss filed Monday, June 8, as part of litigation that had been filed earlier this year by a group of independent artists, songwriters, and producers.

Plaintiffs’ complaint, as described in the reporting, alleges that Lyria 3 was trained on copyrighted songs “ripped” from YouTube without compensating the creators. The plaintiffs seek to hold Google responsible for training an AI system using music uploaded by others, not for generating outputs that directly replicate specific recordings. In response, Google’s lawyers frame the lawsuit as built on assumptions that it trained on the specific works at issue, according to the motion summary.

Billboard reported that Google’s motion does not rest primarily on the company’s position that the training is protected as fair use. Instead, Google says it has permission through the platform’s licensing structure. The filing reportedly characterizes the lawsuit as relying on an “unsupported hypothesis” that Google trained on plaintiffs’ specific works, while also arguing that the relevant permissions flow from the YouTube terms.

The lawsuit falls into a broader wave of AI-related copyright cases brought over training data, including suits involving AI music generators brought by major labels and other creators in recent years. In those cases, defendants have often argued that training can be lawful without explicit licensing if it is transformative or otherwise protected. This case is presented as different because Google owns YouTube, making the terms-of-service argument more central to the dispute.

Separate coverage of the underlying lawsuit reported that the case was filed in March 2026 and that the plaintiffs included independent musicians who also pursued claims in other AI music litigation. That reporting also describes the plaintiffs’ core allegation as training of the Lyria model on music uploaded to YouTube without separate permission for AI purposes.

What happens next depends on the court’s ruling on the motion to dismiss. If the motion is denied, the case would proceed toward further legal briefing and potentially discovery focused on training methods, data handling, and the scope of the rights granted by YouTube’s Terms of Service. If granted, the dispute would likely end at the dismissal stage without reaching a merits determination over copyright infringement, leaving the question of AI training permissions to be litigated in other venues or future cases.

For creators, the stakes include whether uploading to a major platform can be interpreted as authorizing downstream uses for AI training, and whether such permission requires a separate, explicit licensing mechanism. For Google, the key issue is whether contractual terms can shield AI training activities from infringement claims, and how far those terms extend when AI systems are developed using user-uploaded copyrighted works.