Google's Music AI Quietly Feeds on YouTube Creator Content

Google possesses the capability to utilize uploaded content for AI training but remains silent on whether it exercises this option.

For any musician who has uploaded a song to YouTube, it is highly probable that Google views their content as permissible for training its Lyria music AI, a fact the company currently refrains from confirming.

This situation has led to a lawsuit from a collective of independent musicians, who allege that Google unlawfully leveraged their YouTube uploads to train its Lyria 3 model. In response, Google has submitted a motion to dismiss the case, stating:

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. Plaintiffs each granted YouTube, and Google — which provides the service—a broad license to use the uploaded content. That license, present in YouTube’s Terms of Service, authorized the conduct alleged in the Complaint.

This response reflects a conventional legal defense strategy, essentially asserting, "you lack proof of our actions, and even if we did, it falls within our authorized rights." When confronted directly with inquiries regarding the use of YouTube videos to train its Lyria 3 AI music model, Google chose not to comment. Nevertheless, prior public statements strongly suggest an affirmative answer.

Supporting this inference, YouTube CEO Neal Mohan indicated in an April 2024 interview with Bloomberg that "some portion" of YouTube videos could be utilized internally for training models such as Gemini. Further corroboration came later that year via a blog post concerning creator tools, which explicitly stated, "we use content uploaded to YouTube to improve the product experience for creators and viewers across YouTube and Google, including through machine learning and AI applications."

The company additionally confirmed to CNBC its practice of employing YouTube uploads to train Gemini and Veo. However, what Google has consistently avoided is a specific confirmation that YouTube uploads are also being used to train Lyria.

Nevertheless, within its motion to dismiss, Google asserts that by directly uploading content to YouTube, the plaintiffs consented to the terms of service, which explicitly grants the company the right to "reproduce, distribute, [and] prepare derivative works" derived from the uploaded material.

Given these circumstances, Google's reluctance to acknowledge what appears to be an obvious practice might seem perplexing. Yet, at this juncture, the company stands to gain very little by publicly confirming such use. With ongoing litigation, preserving plausible deniability represents a strategic maneuver.