Google has filed a motion to dismiss a copyright lawsuit brought by independent musicians over the training of its Lyria 3 AI music generation model, arguing that artists who uploaded their work to YouTube effectively licensed it for that use under the platform's terms of service. The motion, reported on Music Business Worldwide on June 10, 2026, escalates what had been a pending infringement action into a direct legal confrontation over how broadly platform terms of service can be construed to cover AI training.

The Lawsuit

Independent musicians filed suit against Google earlier this year, alleging that the company mined YouTube's catalog — without obtaining separate licenses — to train Lyria 3, the generative AI music tool Google launched as part of its broader AI creative suite. The plaintiffs contend that uploading a track to YouTube to reach listeners is categorically different from consenting to have that track ingested as training data for a competing commercial AI product. As Billboard reported in March 2026, the core copyright infringement allegation turns on whether Google's use of those sound recordings falls within the scope of any license the artists actually granted — or whether it represents an unauthorized reproduction and derivative use.

The complaint names Google's AI music business practices more broadly alongside the specific Lyria 3 training claims, suggesting the plaintiffs are not limiting their theory to a narrow technical infringement but are challenging the structural relationship between YouTube's content pipeline and Google's AI development stack.

Google's Defense: The Terms of Service Argument

Google's position, as it moved for dismissal, is that YouTube's terms of service already extend to AI model training — that when artists accepted those terms, they granted a license broad enough to cover the company using uploaded content as training material. Billboard's June 9 coverage notes Google framing this as a matter of settled contractual consent rather than copyright infringement at all.

Google has been consistent on the underlying data question since the model launched. When Lyria 3 was released in late March 2026, The Decoder reported that Google stated publicly the model was trained on materials that both YouTube and Google have the right to use under their terms of service and partner agreements. The dismissal motion now translates that public-relations framing into a formal legal argument: there was no infringement because there was no unauthorized use.

The legal weight of that argument hinges almost entirely on what YouTube's terms of service actually say — and, more precisely, what a court decides they can be interpreted to mean. The relevant clause in YouTube's standard upload agreement grants Google a broad, worldwide, royalty-free license to use, reproduce, distribute, and create derivative works from uploaded content for the purpose of operating and improving Google's services. Whether "improving Google's services" extends to training a commercial generative AI product that can itself produce music — potentially displacing the very artists whose tracks fed the model — is the interpretive question the court will now have to address if the dismissal motion fails.

A Pattern the Industry Has Seen Before

This is, in structure, a recognizable playbook. We have seen it run before — most directly in the early years of the commercial internet, when platforms built substantial businesses on user-uploaded content while terms of service gradually expanded to license almost everything short of outright ownership transfer. The music industry fought those expansions repeatedly, from the original Napster litigation through the long battles over the DMCA safe harbor and the "value gap" that eventually prompted the EU's Copyright Directive. Each cycle, the technology moved faster than the legal framework; each cycle, courts and legislators arrived late and imperfectly.

The AI training litigation wave is that dynamic compressed and intensified. The inputs are copyrighted creative works. The outputs are commercially deployed generative systems. The contractual instrument being invoked — a platform terms of service agreement — was almost certainly not drafted with AI training as a contemplated use case when most artists accepted it. Whether courts will hold that a broadly worded license written for one purpose extends by implication to a genuinely novel one is an open question, and the answer will reverberate well beyond Google and this particular set of plaintiffs.

What Is Actually at Stake

The Lyria 3 case sits at the intersection of three distinct legal theories that AI music litigation is beginning to sort out: direct copyright infringement in training (unauthorized reproduction of protected works), derivative work claims (whether a model's outputs infringe the stylistic or compositional character of training inputs), and, separately, the contractual question Google is now pressing — whether platform terms of service can function as a blanket AI training license.

The first two theories are being litigated across multiple jurisdictions and multiple AI modalities simultaneously. The third is newer and, in some ways, more consequential for the broader creator economy. If Google's ToS argument succeeds at the dismissal stage, it establishes a precedent that every major platform — YouTube, SoundCloud, Spotify's podcast network, any service that hosts user content under similarly worded terms — could cite to immunize its AI training operations from copyright claims by the very creators whose work fills those platforms.

For independent musicians specifically, this matters acutely. Major labels have the leverage to negotiate direct licensing deals with AI developers — and several have done so — while independent artists typically have no negotiating position beyond the take-it-or-leave-it terms a platform presents at upload. If those terms are subsequently held to authorize AI training, independent creators will have contributed to the training corpora of commercial generative models without compensation or meaningful consent, and with no structural recourse.

Worth flagging here: even if Google ultimately prevails on the ToS argument, the reputational and policy dynamics around this case are unlikely to track the legal outcome one-to-one. Legislators in the EU and, with less speed, in the US have been watching AI training and copyright with increasing attention. A court ruling that ToS language written in the mid-2010s implicitly covers 2026-era generative AI training would likely accelerate legislative action to close that gap explicitly — as has happened in prior cycles when courts interpreted existing law in ways that left creators without remedies.

Where the Case Goes Next

Google's motion to dismiss is a threshold argument: the company is asking the court to dispose of the case without reaching the merits, on the grounds that the plaintiffs lack a cognizable claim under the existing contractual framework. If the motion is denied, the case proceeds to discovery, at which point the factual record of how Lyria 3 was actually trained — which audio files, in what volumes, under what internal authorization process — would become central. That discovery record, if it materializes, would be informative not just for this case but for the broader litigation landscape.

For the independent artists who filed this action, the dismissal motion is an early but meaningful test. Surviving it keeps the case alive and preserves the possibility of a factual inquiry into Google's training practices. Losing it on ToS grounds would set a significant precedent — one that, in this author's view, deserves legislative scrutiny regardless of how courts rule, given the gap between the circumstances under which those terms were written and the commercial realities of generative AI in 2026.

The next substantive development will be the plaintiffs' opposition brief and the court's ruling on the motion. Both bear watching.