The Lyria fight turns YouTube into both training source and distribution lever

The Verge is pressing the question Google has not clearly answered: whether music uploaded to YouTube is being used to train Lyria. The Kogon et al v. Google lawsuit argues that the issue is not just training data, but a full chain from YouTube to ContentID to AI-generated music.

Google can answer the YouTube data question, but it is leaving it open

The Verge points to the central detail in the Google Lyria dispute: Google is in the best technical and organizational position to know whether music uploaded by creators to YouTube was used to train its music model. From the available text, the company is not giving a clear yes or no.

The related case, Kogon et al v. Google LLC, announced on March 9, 2026 by Loevy + Loevy, makes a broader allegation. The plaintiffs claim Google copied music without authorization, removed copyright management information under the DMCA, trained products such as Lyria 3 and ProducerAI and distributed resulting AI music through its own platforms including YouTube.

Those are allegations, not a court ruling. The important point is the target: vertical integration. Distribution, ownership identification, model training and AI output distribution all sit inside one company.

The platform asymmetry matters more than the model alone

The standard AI copyright debate often stalls on whether training is fair use. This case adds a sharper layer: Google runs the place where musicians upload their work, the ContentID system that identifies it and a product that can generate competing music.

For an independent musician, that is an asymmetry of data and rules. The creator sees a distribution channel. The platform sees a catalog, metadata, audience behavior and possible training material. If that boundary is not defined by contract and technical controls, trust will erode faster than litigation moves.

The business impact goes beyond music. Every marketplace with user-generated content will have to explain whether upload means publishing, moderation, monetization, or training a competing generator too.

The lawsuit has a strong story, but public facts do not settle it

The brake pedal matters here. Loevy + Loevy’s version is a plaintiff narrative and it is rhetorically sharp. The Verge is mostly highlighting Google’s refusal to answer plainly. That is not the same as proving that specific songs entered a specific dataset.

Discovery will decide the useful facts. If internal data pipelines, license agreements, opt-out mechanisms and links between the YouTube catalog and Lyria become visible, the dispute moves from rhetoric to documents.

Logs, licenses and a real right to say no will decide the case

The next signals are practical: whether Google describes Lyria’s data sources, shows licensing coverage, offers a meaningful opt-out and proves ContentID can distinguish original work from AI output that allegedly benefits from it.

For creators, the timing matters too. Rules changed in advance are governance. Rules changed only after lawsuits are damage control. Platforms like to talk about ecosystems, but an ecosystem without clear consent starts to look like a trap.

Lilith's verdict

Google is not only standing at the mixing desk here. It is also at the club door, in the coatroom holding the song catalog and at the register selling music made by a machine.