Google has argued in a recent court filing that YouTube's terms of service grant a broad license for training its Lyria 3 AI music model on user-uploaded songs, according to the source article. The claim, contested by independent artists including Sam Kogon and Magnus Fiennes, could reshape copyright boundaries if accepted by a judge.

The 'worldwide, non-exclusive, royalty-free' clause at the center of the dispute

Google's legal team is hanging its defense on a specific clause in YouTube's terms of service, as reported in the lawsuit documents. The clause grants YouTube a worldwide,non-exclusive, royalty-free, sublicensable, and transferable license to use uploaded content, including for derivative works, in connection with the service and its business. Google argues that this language explicitly covers AI training — a reading that would give it a free pass to train models like Lyria 3 on nearly any music uploaded to the platform.

The company's approach deliberately sidesteps fair use, the defense most commonly invoked by AI firms facing copyright lawsuits. Instead, it asserts an express contractual license, which could be harder for plaintiffs to overcome if the court agrees the clause is unambiguous.

Why Sam Kogon, Magnus Fiennes and other artists are pushing back

The plaintiffs — a group of independent artists, songwriters, and producers — reject Google's interpretation outright. According to the source, they argue that the license does not extend to AI training and that their works were used without permission. Kogon and Fiennes are among those leading the charge, and they have signaled they will oppose Google's motion to dismiss.

For these creators, the stakes are personal: their music, uploaded to YouTube for public display, is now being used to train a commercial AI model without any additional compensation or consent. Their case highlights the growing chasm between how users understood platform terms at the time of upload and how companies interpret them years later.

Warner Music Group’s licensed deals vs. Universal’s Suno lawsuit

The legal landscape is further complicated by the fact that major labels have already struk separate deals with AI companies. The source notes that Warner Music Group has engaged in licensed training agreements, while Universal Music Group continues to sue the AI startup Suno. Those label-specific contracts sometimes include explicit AI provisions, which Google's blanket terms-of-service argument does not address.

This patchwork of agreements means that even if Google prevails on its terms-of-service claim for user-uploaded content, the same logic may not apply to music distributed under those separate licensing frameworks.. The court may have to weigh whether a platform's standard terms can override or coexist with individual label deals.

If Google wins, every YouTube upload becomes AI training data

The broader implication of this case extends far beyond the Lyria 3 model. Should the court accept Google's interpretation,it would set a precedent that any content uploaded to YouTube — videos, audio, commentary — is fair game for training AI systems under the existing terms of service. that would effectively give Google and parent company Alphabet a massive, royalty-free dataset built from user contributions over the platform's entire history.

As the source article observes, this case is part of a larger wave of litigation where AI companies face accusations of illegal training. But Google's contract-based defense is novel, and if it succeeds, it could embolden other platforms with similar broad-rights clauses to follow suit.

The unanswered question: Can a 2010-era clickwrap bind a 2024 AI model?

A central open question is whether a terms-of-service agreement written years before generative AI existed can reasonably be interpreted to authorize such use.. The plaintiffs are expected to argue that no user in 2015 or even 2020 could have understood that clicking “I agree” granted permission to train a commercial AI music generator. The source does not provide Google's response to that argument in detail,leaving the court to decide how far a general-purpose license can stretch.

Also unaddressed in the source is whether YouTube's content moderation and takedown policies — which allow creators to remove their videos — would supersede the license grant if a user later objects to AI training. These ambiguities ensure that the case will be closely watched by both copyright lawyers and AI developers.