Google won’t just admit it’s feeding YouTube creators to its music AI
Google is refusing to confirm whether it used YouTube uploads to train its Lyria music model, even as it argues a lawsuit over that practice should be dismissed.
Intelligence analysis by GPT-5.4 Mini
A group of independent musicians says Google trained its Lyria 3 music model on songs they uploaded to YouTube. Google won’t confirm the training data, but in court it argues the creators granted broad rights through YouTube’s terms of service.
Google is being accused of using songs people posted on YouTube to teach its music-making AI, like showing a robot lots of songs so it can learn the pattern. Google won’t say if that happened, and it says the upload rules may have allowed it anyway.
Analysis
What happened
A group of independent musicians is suing Google, claiming the company illegally used songs they uploaded to YouTube to train its Lyria 3 music model. Google filed a motion to dismiss, arguing the lawsuit rests on an unproven assumption that it trained on those specific works and that the complaint cannot stand even if the allegations are treated as true.
Google’s defense
The company’s main legal point is that creators agreed to YouTube’s terms of service when they uploaded their content. Google says those terms gave it a broad license to use the material, including rights to reproduce, distribute, and prepare derivative works based on the uploads. In other words, Google is arguing that the uploads were covered by the contract creators accepted.
Why the refusal to answer matters
When asked directly whether YouTube videos were used to train Lyria, Google declined to comment. The Verge notes that the company has publicly acknowledged using YouTube content for other AI systems: YouTube CEO Neal Mohan said in 2024 that some YouTube videos may be used internally to train models like Gemini, and later Google said YouTube uploads are used to improve products through machine learning and AI applications. Google also confirmed to CNBC that it used YouTube uploads to train Gemini and Veo.
What Google has not said is that Lyria specifically used YouTube uploads. The article’s framing is that the company has little incentive to admit that detail while litigation is pending, because keeping some ambiguity may be strategically useful in court.
Key points
- Independent musicians are suing Google over claims that their YouTube uploads were used to train the Lyria 3 music model.
- Google asked the court to dismiss the case, arguing the plaintiffs agreed to YouTube’s broad licensing terms.
- When asked directly whether Lyria used YouTube videos, Google declined to comment.
- The Verge points to Google’s earlier public statements showing that YouTube uploads have been used for other AI training and product improvement.
- The dispute highlights the gap between platform terms and creators’ expectations about AI training.
If the case forces clearer disclosure, creators could get a better understanding of how their uploads are used across Google’s AI systems. It could also push platforms to spell out training rights more explicitly, which would make the rules easier to judge before artists upload their work.
If Google’s position holds, creators may find that uploading work to a platform gives the company broad room to use it for AI training. That could deepen distrust between artists and major platforms, especially if companies avoid being specific about which models use creator content.
