YouTube Terms of Service Allow AI Music Training, Google Says
AI
YouTube
Google
TLDR: CHICAGOāGoogle says YouTube terms let it train AI music models on uploaded songs without extra permission, derailing an indie copyright suit over Lyria 3.
Key Takeaways:
- Background includes a wave of AI training lawsuits over copyrighted music, text, and images, with fair use defenses common.
- Google files a motion to dismiss, citing YouTubeās terms that grant YouTube a royalty free, sublicensable license to use uploaded content.
- If the court accepts Googleās license argument, Google could treat direct uploads as authorized training data, while label licenses may still limit AI rights.
- Plaintiffs name Sam Kogon, Magnus Fiennes, Michael Mell, Attack the Sound, Stan Burjek and James Burjek, plus Directrix.
- Dataset contents for Lyria 3 remain undisclosed, and the next step is opposition filings before the motion is decided.
Google is trying to move the fight from copyright theory to contract fine print. If the court agrees, indie artists may face a new obstacle: the platform rulebook, not the data itself.
Google is trying to move the fight from copyright theory to contract fine print. If the court agrees, indie artists may face a new obstacle: the platform rulebook, not the data itself.
Q&A
If Google wins on YouTube direct uploads, what will plaintiffs target next in similar AI music cases?
They will likely focus on uploads that fall outside YouTubeās user terms, plus disputes over scope like derivative works rights and how training extracts or transforms audio.
How might the presence of label and publisher licensing deals change the outcome for future AI training claims?
Courts may treat those contracts as a separate permission pathway, where derivative works language and AI specific clauses control what training is allowed.
Why does Googleās ownership of YouTube make this case harder to frame as pure fair use?
Because Google can point to internal service permissions granted by users, reducing the need to prove ātransformativeā use under fair use.
What evidentiary pressure does Googleās refusal to disclose the Lyria 3 dataset create for both sides?
Indies may push for discovery to map which tracks were used, while Google will argue allegations are speculative without concrete proof tied to specific works.
Could this motion to dismiss influence how major labels negotiate āguardrailsā for Gen AI across platforms?
Yes, it can drive licensing language toward clearer AI training boundaries, including whether training is allowed, where outputs can be used, and whether any royalties apply.
No comments yet. Be the first to share your thoughts!