Google says YouTube music uploads can be used to train AI in Lyria lawsuit defence
Google claims its YouTube terms already grant the rights needed for Lyria 3 AI training.
Google is seeking to dismiss a copyright lawsuit brought by independent musicians over its Lyria 3 AI music model, arguing that music uploaded to YouTube is already licensed for uses that include AI training. The claim is part of a broader legal dispute over how far platform terms extend when it comes to generative AI.
The case, reported by Music Business Worldwide, centres on allegations that Google trained Lyria 3 on copyrighted recordings without permission. But Google’s defence hinges on a different interpretation. The company insists that artists granted permission when they uploaded their music to YouTube in the first place.
Google’s defence – YouTube terms already cover AI training
In its court filing, Google argues that YouTube’s Terms of Service give the platform and its “affiliates” a broad, royalty-free licence to use uploaded content in multiple ways.
As reported by Complete Music Update, Google says this includes the right to “reproduce, distribute and prepare derivative works”. Language, it believes, is wide enough to cover AI model training.
The company also points to the structure of Alphabet, noting that “affiliates” includes other divisions of Google itself. On that basis, it argues that Lyria 3 falls within the scope of rights already granted by users when they upload content to the platform.
Google further highlights the scale of its existing music ecosystem, including more than $8 billion paid to the music industry via YouTube over a 12-month period, positioning its AI development as part of the same broader platform activity.
Independent artists say AI training was never part of the deal
The musicians behind the lawsuit dispute that interpretation, arguing that uploading music for streaming and monetisation is not the same as granting permission for AI model training.
As covered by Digital Music News, the artists allege that their copyrighted recordings were used without authorisation to train Lyria 3, and that YouTube’s ownership structure gives Google additional leverage through systems like Content ID.
They also argue that there is no transparency around whether their specific works were used at all, meaning the claim rests on inference rather than confirmed evidence.
Google, however, counters that the lawsuit is built on what it calls an “unsupported hypothesis” that any of the plaintiffs’ recordings were actually included in training data.
Lyria 3 Pro and the expanding AI music push
This legal dispute comes at a time when Google is rapidly expanding its AI music tools. As we previously reported, Lyria 3 Pro represents a significant step forward in generative music technology, moving beyond short audio snippets into more structured, longer-form compositions.
The model can now generate tracks of up to three minutes and allows users to specify elements like intros, verses, and choruses. It is also being integrated across Google’s ecosystem, including Gemini and other AI tools.
Google says its approach to training data remains responsible, stating that it only uses materials it has the right to access under platform terms and partner agreements. It also claims safeguards are in place to prevent imitation of specific artists, alongside watermarking systems like SynthID.
What this means for independent artists
For independent musicians, the case highlights a growing gap between how platform terms are written and how creators interpret them. Uploading music to YouTube remains essential for discovery and income, but the legal rights attached to those uploads are increasingly under scrutiny as AI systems evolve.
If Google’s argument succeeds, it could reinforce the idea that existing platform licences already cover AI training use cases. If it fails, it may force platforms to rethink how AI development is separated from standard content licensing.
Either way, the outcome is likely to influence how digital distributors, streaming platforms, and AI developers define rights around music data going forward.
AI is moving faster than the law
Beyond the courtroom, the situation reflects a wider industry shift. AI-generated music is rapidly moving from short-form experimentation into full track creation, increasing the volume of synthetic content entering the market.
For artists, that raises ongoing questions around competition, originality, and how their work may be used within training datasets they have limited visibility over. As AI tools like Lyria 3 Pro continue to evolve, the tension between innovation and consent is only likely to intensify.