YouTube Terms of Service Allow AI Music Training, Google Says in Copyright Lawsuit

Google says in a new court filing that YouTube’s terms of service grant a “broad license” for artificial intelligence models to be trained on music uploaded directly to the platform.

This argument came in Google’s Monday (June 8) motion to dismiss copyright infringement litigation filed earlier this year by a group of independent artists, songwriters and producers. The lawsuit claims that Lyria 3, the AI music model launched by Google in February, was trained on songs ripped from YouTube without compensating artists.

The case is among dozens that have accused AI companies of illegally training their models on copyrighted sound, text and images in recent years. In most of these lawsuits — including ones brought by the major record labels against AI music generators Suno and Udio — tech companies have argued that the so-called fair use principle of copyright law allows them to create transformative works without licensing the training data.

The Google case is different, though, because the tech giant owns YouTube, one of the biggest music streaming platforms in the world. Rather than making a fair use argument, Google’s lawyers say in Monday’s brief that they actually do have a license to train AI with the songs uploaded to YouTube by these indie artists.

“Their lawsuit is based on the unsupported hypothesis that Google trained on their specific works,” reads the motion, filed by powerhouse litigation firm Quinn Emanuel. “Even accepting their untested allegations as fact, the complaint cannot stand. Plaintiffs each granted YouTube, and Google — which provides the service — a broad license to use the uploaded content. That license, present in YouTube’s terms of service, authorized the conduct alleged in the complaint.”

Specifically, the motion cites a clause in YouTube’s terms of service that states, “By providing content to the service, you grant to YouTube a worldwide, non-exclusive, royalty-free, sublicensable and transferable license to use that content (including to reproduce, distribute, prepare derivative works, display and perform it) in connection with the service and YouTube’s (and its successors’ and affiliates’) business.”

These terms of service govern all individual users who upload music directly to YouTube – meaning, if Google’s argument is accepted by the court, the company would be legally permitted to train AI models on and otherwise utilize that content for “derivative works.” However, the situation might be different for music distributed through YouTube pursuant to licensing agreements with labels and publishers.

All of the major music companies, as well as indie labels and publishers, have licensing deals with YouTube that govern copyright protection and royalty payouts. Whether AI training is permitted likely differs from license to license, depending on the grants around derivative works. Some companies have even added AI-specific language in recent years that sets forth rules for whether and how music on the platform can be used for Google’s AI initiatives.

For example, during the investor earnings call for Universal Music Group (UMG) this past October, CEO Lucian Grainge said the company’s renewed licensing deal with YouTube “secured really important guardrails and protection for our artists and writers around Gen AI content.”

Google has not explicitly disclosed what material was included in the Lyria 3 training dataset. In a statement to Billboard when the model was rolled out, a Google representative said it was trained on music that Google has “a right to use under our terms of service, partner agreements, and applicable law.”

A rep for Google declined on Tuesday (June 9) to comment beyond the court papers. The attorneys pursuing the Lyria 3 lawsuit did not immediately return a request for comment. They will file opposition papers fighting Google’s motion to dismiss the case in the coming weeks.

As a proposed class action, the lawsuit is aimed at eventually representing all indie artists who’ve allegedly been harmed by Google’s AI training. Its named plaintiffs are singer/songwriter Sam Kogon, composer/producer Magnus Fiennes, songwriter/producer Michael Mell, R&B group Attack the Sound, father-and-son folk rock duo Stan Burjek and James Burjek, and the Chicago-based band Directrix.

This same group previously brought similar copyright lawsuits against AI music companies Suno, Udio and Mureka. The more closely watched front in the AI music copyright war, though, is litigation launched in 2024 by UMG, Warner Music Group (WMG) and Sony alleging Suno and Udio infringed their IP “at an almost unimaginable scale.”

UMG and WMG both settled with Udio in the fall, striking deals for the company to train a new model on licensed work that will remain within a “walled garden” on the platform. WMG later settled with Suno as well, agreeing to a licensed training model that allows for off-platform distribution. UMG is still suing Suno, and Sony has not settled with either company.