Google won't just admit it's feeding YouTube creators to its music AI

Overview

A group of independent musicians is suing Google, claiming the company trained its Lyria 3 music-generation AI on their songs uploaded to YouTube without permission or pay. Google has filed a motion to dismiss, arguing that YouTube's terms of service already grant it a broad license to use those uploads, and that the musicians have not proven their specific works were used. Notably, Google has not directly confirmed whether it trained Lyria on the songs in question. The proposed class action sits in the US District Court for the Northern District of Illinois.

Key Takeaways

Independent musicians allege Google trained its Lyria 3 model on roughly 44 million audio clips scraped from YouTube, about 280,000 hours of music, without consent or compensation.
Google's defense leans on YouTube's terms of service, which it says grant a broad license to reproduce and create derivative works from anything users upload.
Google's motion to dismiss also argues the plaintiffs cannot prove their specific songs were part of the training data, calling the claim an unsupported hypothesis.
The lawsuit goes beyond copyright, adding claims under the DMCA, the Lanham Act for false endorsement, and a biometric privacy law tied to alleged extraction of artists' voiceprints.
Named plaintiffs include singer Sam Kogon, composer Magnus Fiennes, producer Michael Mell, and the Chicago band Directrix, with the suit seeking class-action status for other indie artists.
The case tests whether agreeing to a streaming platform's upload terms also means consenting to having your work feed a commercial AI model.

Stats & Key Facts

#Roughly 44 million audio clips were allegedly scraped from YouTube to train Lyria 3.
#The training set is estimated at about 280,000 hours of music.
#Lyria 3 generates tracks up to about 3 minutes long, complete with vocals and lyrics.
#The proposed class action names a single defendant, Google, in the Northern District of Illinois.
#At least six named plaintiff acts are listed, including solo artists, a duo, and full bands.
#The original complaint was filed on March 6, 2026, with Google's motion to dismiss following in June 2026.

Google won't just admit it's feeding YouTube creators to its music AI

What independent musicians accuse Google of doing with Lyria 3

The heart of the case is a claim that Google used artists' work to build a competing product.

A group of independent musicians and songwriters filed suit against Google in March 2026, accusing the company of training its Lyria 3 music AI on songs they had posted to YouTube. The complaint estimates that Google scraped about 44 million audio clips, roughly 280,000 hours of music, and fed them into the model without asking permission or paying the people who made the recordings.

The plaintiffs frame this as plain copyright infringement: their creative work went into a machine learning system that now produces music, and they were never consulted. The suit is structured as a proposed class action, meaning the named artists want to represent a much larger pool of independent creators who say they were affected the same way.

How Google's motion to dismiss leans on YouTube's upload license

Google's main defense is a contract argument, not a denial.

Rather than directly confirming or denying that it trained on these specific songs, Google asked the court to throw the case out. Its central argument points to YouTube's terms of service, which every uploader agrees to. Google says those terms grant it a worldwide, non-exclusive, royalty-free, sublicensable and transferable license to use uploaded content, including the right to reproduce, distribute, and prepare derivative works.

In Google's reading, training an AI model falls within that license. The company essentially argues that the musicians signed away the rights they are now suing over the moment they clicked to accept YouTube's terms and posted their tracks.

Why Google says the plaintiffs cannot prove their case

A second prong of the defense attacks the evidence itself.

›Google calls the suit built on an unsupported hypothesis that it trained on the artists' specific works.
›The company argues that even if the allegations are accepted as true, the complaint still fails as a legal matter.
›Google contends the plaintiffs lack standing to bring some of the claims, including those under the Digital Millennium Copyright Act.
›It disputes that the artists have voices distinctive enough to support trademark-style false endorsement claims.

The broader set of legal claims beyond copyright

The musicians stacked several theories on top of standard infringement.

›Copyright infringement for training the model on protected recordings.
›Violations of the Digital Millennium Copyright Act tied to handling of rights-management information.
›Contributory infringement, alleging Google induced or enabled further copying.
›False endorsement under the Lanham Act, claiming the AI trades on artists' identities.
›Biometric privacy violations, alleging Google extracted artists' voiceprints from their recordings.

Who the named plaintiffs are

The case is led by working artists rather than major labels.

›Sam Kogon, a New York singer-songwriter.
›Magnus Fiennes, a Los Angeles composer and producer.
›Michael Mell, a songwriter and producer.
›Attack the Sound, an R&B group.
›Stan Burjek and James Burjek, a father-and-son folk rock duo.
›Directrix, a Chicago-based band.

What the fight means for anyone who posts work online

The outcome reaches well past music.

The core question is whether agreeing to a platform's standard upload terms also means agreeing to let that platform feed your work into a commercial AI system. Google's position suggests that the broad licenses common across user-generated content sites might cover AI training. The musicians counter that uploading a song for streaming and monetization is a different thing from handing over training rights for a product that competes with them.

For non-technical creators and business owners, the takeaway is practical. The terms of service you accept on video, photo, and writing platforms often contain wide-reaching license language, and courts are now testing how far that language stretches in the age of generative AI. A ruling here could shape how every uploader thinks about what happens to their content next.

Where the case goes from here

The dispute is early and far from settled.

Google's motion to dismiss is the company's opening move, and the plaintiffs' attorneys are expected to file opposition papers fighting it. A judge will then decide whether the case proceeds toward discovery, where Google might be pressed to reveal what actually went into Lyria 3's training data.

The case is one of a growing wave of disputes over how AI companies source the material that powers their models. Its result could influence other suits brought by writers, visual artists, and performers who say their work was used without consent.

Frequently Asked Questions

What is Lyria 3?

Lyria 3 is Google's music-generation AI, available inside its Gemini app, that produces full tracks with vocals and lyrics up to about three minutes long from text prompts. The lawsuit centers on what data was used to train it.

What are the musicians actually claiming?

They allege Google scraped roughly 44 million audio clips from YouTube, about 280,000 hours of music, and used them to train Lyria 3 without consent or payment. Their claims include copyright infringement, DMCA violations, false endorsement, and biometric privacy violations.

What is Google's defense?

Google argues that YouTube's terms of service already grant it a broad license to reproduce and create derivative works from uploaded content, so the artists licensed the use when they posted their songs. It also says the plaintiffs cannot prove their specific works were in the training data.

Does uploading to YouTube really let Google train AI on my music?

That is the unsettled legal question at the center of this case. Google says its upload license covers such use, while the musicians argue that posting for streaming is not the same as consenting to AI training, and a court has not yet decided who is right.

Where and when was the case filed?

The suit was filed in March 2026 in the US District Court for the Northern District of Illinois as a proposed class action, with Google as the sole defendant. Google filed its motion to dismiss in June 2026.

The dispute turns on a single unsettled question: whether clicking to accept a platform's upload terms also hands over the right to train commercial AI on your work. Until a judge rules on Google's motion, that question stays open for every creator who posts online.