Google Says Musicians Already Agreed to AI Training. Here's Why That Matters.

Google has asked a court to dismiss a lawsuit brought by independent musicians who claim the company used their songs to train an AI music generator without permission. Google's argument is straightforward: when these artists uploaded their music to YouTube, they already agreed to let the company use it in whatever way the platform's terms of service allow. This disagreement is now shaping what could become a much larger legal battle over how AI developers can use creative work.

What the Lawsuit Claims

Earlier this year, independent musicians sued Google over Lyria 3, an AI tool that can generate music. They claim Google took their songs from YouTube — where they had uploaded them to reach listeners — and used them to train this AI system without getting separate permission or paying them. According to Billboard, the musicians say uploading a song to find an audience is completely different from allowing that song to train a competing AI product that could eventually replace musicians' work.

The lawsuit targets Google's broader AI music practices, not just this one case. The musicians are arguing that Google's entire approach — mining YouTube's catalog to build AI models — is the real problem.

Google's Defense

Google's legal team says the answer is already written in the fine print. When people upload videos or music to YouTube, they click "agree" to the site's terms of service. Those terms say Google can use uploaded content to "operate and improve" its services. Google contends that training an AI music generator falls within that permission — that no extra agreement was needed.

When Google first launched Lyria 3 in March 2026, The Decoder reported that the company stated publicly it had the right to use the training material. Now, in court, Google is making that same claim as a legal argument: no permission was broken because no permission was needed in the first place.

The whole case comes down to one interpretive question: does a terms of service agreement written years ago — before AI music generation existed — actually give Google permission to build an AI system that might compete with the very musicians whose songs trained it?

Why This Pattern Looks Familiar

This is not the first time we have seen this movie. During the early days of the internet, platforms built large businesses on content users uploaded while slowly expanding what the terms of service allowed. The music industry fought back — through Napster lawsuits, battles over download regulations, and disputes over how much payment streaming services owed. Each time, the technology moved faster than the courts could rule, and the law caught up late and imperfectly.

What is happening with AI is the same dynamic, but faster and more intense. Copyrighted songs are fed into AI systems. Those systems generate commercial products. But the contract being used to justify it was written for a very different purpose, when neither the company nor the artist was thinking about AI training at all.

What Hangs in the Balance

Three different legal questions are starting to separate in AI music cases across the country. The first is whether Google simply copied songs without permission. The second is whether the songs the AI produces sound too much like the originals. The third — which Google is pushing in this case — is whether a terms of service contract can act as a blanket license for AI training.

That third question matters most, because if Google wins, every major platform hosting user content — YouTube, TikTok, Spotify's podcast service, SoundCloud — could use the same argument to protect their AI training from copyright lawsuits.

Independent musicians are in a particularly tight spot. Major record labels can sit down and negotiate their own AI licensing deals with Google and other companies. Independent artists cannot. They have no bargaining power. They either accept whatever the platform's terms say, or they do not upload at all. If courts decide those old terms already covered AI training, independent creators will have fed commercial AI systems with no compensation, no real choice, and no legal recourse.

The broader policy context here deserves attention. Even if Google wins this case on purely legal grounds, legislators in Europe and the United States have been paying close attention to how AI systems are trained on copyrighted work. A court ruling that says old terms of service implicitly cover new AI technology would very likely push lawmakers to write new laws that explicitly settle the question — the way previous technology cycles have shown us they would.

What Happens Next

Google is asking the court to stop the case before it even gets started, claiming the musicians have no valid legal claim. If the judge agrees with Google, the lawsuit ends. If the judge disagrees, the case moves forward, and both sides can examine what actually happened — which songs were used, how much of each, what Google's internal rules were for training the system. That detailed factual record would matter not just for this case but for all the other AI training copyright cases working through the courts.

For these independent musicians, surviving this dismissal motion is the first hurdle. If they clear it, the case stays alive and they get a chance to show what Google actually did. If Google wins this round on the terms of service argument, it would set a significant precedent that other platforms could use to shield their AI training from challenge. Whether that is the right outcome is, in my view, something lawmakers need to address directly, regardless of how courts rule — because the gap between when those terms were written and how AI actually works today is simply too wide for old contract language to fairly settle.