Google has asked a court to dismiss a lawsuit from a group of independent musicians who allege the company trained its Lyria 3 music model on songs they uploaded to YouTube. The motion, filed in connection with a June 10, 2026 case, does not deny the training took place. Instead, it argues YouTube's Terms of Service already authorized whatever happened.
The legal strategy is two-pronged. First, Google says the plaintiffs cannot prove their specific tracks were used. Second, even if they were, the upload license they agreed to covers it.
From the motion: the lawsuit rests on an "unsupported hypothesis that Google trained on their specific works," and the complaint "cannot stand" because plaintiffs granted YouTube and Google "a broad license to use the uploaded content" — a license that, Google says, "authorized the conduct alleged in the Complaint."
“Their lawsuit is based on the unsupported hypothesis that Google trained on their specific works. Even accepting their untested allegations as fact, the Complaint cannot stand.”— Google, from its motion to dismiss
Key facts
- 01A group of independent musicians sued Google in 2026, alleging it trained its Lyria 3 music model on songs they uploaded to YouTube.
- 02Google filed a motion to dismiss on June 10, 2026, arguing YouTube's Terms of Service grant it a broad license to reproduce and prepare derivative works.
- 03YouTube CEO Neal Mohan told Bloomberg in April 2024 that 'some portion' of YouTube videos may be used to train models like Gemini.
- 04Google has publicly confirmed YouTube uploads are used to train Gemini and Veo, but has not specifically confirmed the same for Lyria.
Asked directly whether YouTube uploads feed Lyria 3, Google declined to comment. The musicians have not yet had the opportunity to respond to the dismissal motion in court, and discovery — which would determine what was actually in the training set — has not occurred.
What makes the silence notable is that Google has been comparatively open about training other models on YouTube content. The company has confirmed that uploads are used to train Gemini and its Veo video model. It has not extended that confirmation to Lyria.
YouTube CEO Neal Mohan told Bloomberg in April 2024 that "some portion" of YouTube videos may be used internally to train models like Gemini. A Google blog post later that year went further, stating the company uses "content uploaded to YouTube to improve the product experience for creators and viewers across YouTube and Google, including through machine learning and AI applications."
Lyria 3 is Google's flagship music generation model, capable of producing full instrumental and vocal tracks from text prompts. It powers music features inside YouTube's creator tooling and Google's broader generative stack. A model that produces music plausibly needs music to learn from, and Google sits on the largest repository of user-uploaded audio in the world.
The terms-of-service argument is the more interesting half of the filing. YouTube's upload license grants Google the right to "reproduce, distribute, [and] prepare derivative works" based on user content. Google is effectively arguing that AI training falls inside that grant. If a court agrees, the precedent would extend to every other platform with a similar boilerplate license — which is most of them.
The plaintiffs will argue the license was not contemplated to cover generative AI training when they agreed to it, and that derivative-works language refers to operational uses like transcoding and thumbnailing, not building a competing music product. That dispute — whether platform terms drafted before generative AI silently authorized it — is now the central question in multiple ongoing cases against AI providers.
The case sits alongside Warner Music's recent acquisition of Sureel AI, which we covered earlier, as the music industry hunts for both legal remedies and technical ones to track how its catalog ends up inside models. Independent musicians without a major label's litigation budget have fewer options, which is partly why a class-style case against Google is being tested now.
Google's calculation is straightforward. Confirming Lyria's training data publicly while litigation is pending gives plaintiffs ammunition and invites copycat suits from labels, publishers, and individual artists across YouTube's catalog. Declining to comment costs nothing and preserves negotiating room. The motion to dismiss is the legal equivalent of running out the clock until the terms-of-service question is resolved on Google's preferred ground.
The outcome will matter well beyond Lyria. If platform terms of service are deemed sufficient to license generative AI training on user content, the largest content hosts — YouTube, Meta, Reddit, X — gain a structural moat over every AI lab that has to license data the old-fashioned way. If they aren't, Google's music model and a good chunk of the rest of its generative stack will need a retroactive licensing story it has so far avoided telling.
Working on something we should cover, or seeing a story we missed? Send leads, documents, or feedback to hello@aichatdaily.com. For sensitive tips, see our secure tips page for Signal and PGP options.
Spotted an error? Email hello@aichatdaily.com with the URL and the issue, or read our full corrections policy.
