Google isn’t just admitting that it’s feeding YouTube creators to its music AI

If you’ve uploaded music to YouTube, Google also considers your video games good for music training for Lyria AI, it won’t accept it right now.

Their case is based on the inconsistent opinion that Google taught on their actual work. Even accepting their untested claims as truth, the Complaints will not stand. The defendants each granted YouTube, and Google – which provides the service – permission to use the uploaded content. The license, found in YouTube’s Terms of Service, acknowledged the complaint.

This is a common stumbling block in legal transfers: “you can’t prove we’ve done it, and even if we did, we’re allowed to do it.” When asked directly if Google is using YouTube videos to train its Lyria 3 AI music model, the company declined to comment. But based on what people have said in the past, it seems safe to assume yes.

In an interview with April 2024 and BloombergYouTube CEO Neal Mohan said that “some portion” of YouTube videos can be used internally to train models like Gemini. Later that year, a blog post of production tools was confirmed, saying that “we use the content uploaded on YouTube to improve the content of creators and viewers on YouTube and Google, including machine learning and AI programs.”

The company also confirmed that Price CNBC that it was using YouTube uploads to train Gemini and Veo. What Google hasn’t done is confirm that it’s also using YouTube’s uploads to train Lyria.

However, in order to deny, Google says that by uploading the content directly to YouTube, the plaintiffs agreed to the terms of use, which gives the company the right to “produce, distribute, (and) prepare derivative works,” based on the uploaded content.

With all of this in mind, it might seem counterintuitive that Google doesn’t accept the obvious. But right now, the company has little value in writing history. And while the lawsuits are pending, maintaining a valid denial is a calculated move.