Uploaded songs on YouTube? Google says you gave permission to use it to train AI

In a court filing, Google claims that artists who uploaded music on YouTube gave permission to use the music to train AI as per the platform's terms of service. Google is currently facing a lawsuit by a group of independent artists who allege that their music was used to train Google's Lyria 3 audio model.

In Short

• Google says songs uploaded on YouTube can be used to train AI
• The company says independent artists agreed to it via YouTube’s terms of service
• Google is being sued for allegedly using songs from YouTube to train Lyria 3 AI model

In February this year, Google launched Lyria 3 through Gemini. The company claimed that Lyria 3 could generate songs for you based on prompts. But shortly after its release, a lawsuit was filed against Google for allegedly training the model on songs taken from YouTube without compensation. And now, in a new court filing, Google has said that by uploading songs on YouTube, artists gave it a broad license to train AI on their music.

As per court documents, Google says that YouTube’s terms of service give it a “broad licensee” to use music uploaded directly to the platform, including for training artificial intelligence models such as Lyria 3. That is, if you uploaded a song on YouTube, the world’s biggest video streaming platform, you gave Google the permission to potentially use it to train AI models.

The copyright infringement lawsuit was filed in March this year by a group of independent artists, songwriters and producers.

Google doesn’t confirm if it actually used music to train AI

One thing to note here is that Google’s case rests on an “unsupported hypothesis.” That is, the tech giant is not saying that it actually used music uploaded on YouTube to train its AI models.

Google’s motion, filed by litigation firm Quinn Emanuel reads, “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, authorised the conduct alleged in the complaint.”

In other words, Google claims that even if it were to use the uploaded content to train AI models, it already received authorisation from the creators when they accepted YouTube’s terms of service.

The filing cites the clause, “By providing content to the service, you grant to YouTube a worldwide, non-exclusive, royalty-free, sublicensable and transferable license to use that content (including to reproduce, distribute, prepare derivative works, display and perform it) in connection with the service and YouTube’s (and its successors’ and affiliates’) business.”

The position marks a different legal approach from the one taken by other AI companies facing copyright claims. In cases involving Suno, Udio and Anthropic, the companies have mainly argued that training on copyrighted material is protected by fair use because the resulting systems are transformative.

Google’s argument in this case is that it already has a licence through YouTube for music uploaded by individual users. If that argument is accepted, it could allow the company to train AI systems on such uploads and use them for derivative works.

This lawsuit alleges that Google’s conduct goes beyond the music generator itself because of the “structural leverage” it holds through ownership of YouTube and Content ID. “Google didn’t just have access to Plaintiffs’ music; it operated the infrastructure through which much of that music reached the world,” the complaint said.

Google is also facing a lawsuit from a group of journalists, podcasters and audiobook narrator for allegedly using their voices to train AI models.

Can Google use all songs on YouTube to train AI?

Keep in mind that Google’s position may be different for music distributed to YouTube under separate licensing agreements with labels and publishers. Major music companies, as well as independent labels and publishers, have licensing deals with YouTube that govern copyright protection and royalty payments, and the treatment of AI training may vary from one agreement to another.

As per reports, some agreements now include AI-specific terms. During an investor call in October, Universal Music Group chief executive Lucian Grainge said the company’s renewed deal with YouTube had secured “really important guardrails and protection for our artists and writers around Gen AI content."

Google has not publicly disclosed the material used to train Lyria 3. When the model was introduced, a company representative said it had been trained on music that Google had “a right to use under our terms of service, partner agreements, and applicable law.”

- Ends