YouTube has long been one of the most accessible ways for independent artists to get their music out into the world: Anyone can create an account and post content on the site with just a few clicks. But what many artists likely didn’t realize when they clicked “agree” to the platform’s terms of service is that YouTube, and its parent company Google, would later claim the agreement justifies training artificial intelligence models on their music.
Google revealed this position in a legal filing earlier this month, obtained and reported by Billboard, as part of copyright litigation brought by indie artists over the training of its AI music model Lyria 3. While Google did not say whether the artists’ music from YouTube was in the Lyria 3 training data set, it argued that this theoretically would be allowed because the YouTube terms of service grant a “broad license to use the uploaded content” as training fodder.
Related
YouTube Terms of Service Allow AI Music Training, Google Says in Copyright Lawsuit
Jack White's Wife Olivia Jean Files for Divorce After 4 Years of Marriage
Kazumasa Izawa of JASRAC on Meeting the New Challenges of Japanese Music Worldwide: Billboard Global Power Players Interview
This is markedly different from the argument made in court by companies behind other AI music models, such as Suno, that they should be free to train on unlicensed music ripped from the internet due to the fair use principle of copyright law. That’s because Google holds a distinctive position, as the owner of one of the world’s largest music streaming platforms, to make a novel legal argument that it actually does have a license to use everything on YouTube — thus skirting the fair use question altogether.
Google’s new argument has quickly raised alarm bells among artist advocates, including Ian Harrison, CEO of the American Association of Independent Music (A2IM). “The use of anyone’s creative work should be consented to very clearly, the terms should be laid out and the value should be shared with the creators,” Harding tells Billboard. “That’s what’s so concerning about this issue [with Google]. It’s really the opposite of that.”
Ron Gubitz, executive director of the Music Artists Coalition (MAC), is concerned that YouTube’s terms of service are too “generic” to properly put musicians on notice that their uploads could be used for AI training. Looking at the fine print, this contract states that YouTube and its affiliates have a worldwide, royalty-free license to use content on the platform to “reproduce, distribute [and] prepare derivative works.” The words “artificial intelligence” are not present, nor is the term “training.” And the terms of service cited by Google are from 2019, years before generative AI came to market.
“Our take is this that this is not informed consent for AI training or output,” says Gubitz. “It was not written with AI training in mind. Consent should be specific and forward-looking, not using a checkbox from however long ago, because that was not the artists’ intent. They were not agreeing to have their music and videos provided as training data for something that could be a competitor for them.”
YouTube’s AI training argument is specifically making waves in the indie music community because the platform’s terms of service apply only to user-generated content (UGC) — that is, music uploaded individually by self-represented artists directly to the site.
Related
Google’s Gemini Launches New AI Music Model, Lyria 3
Harrison tells Billboard he finds it “troubling” that YouTube could be singling out the music of indie creators for broad AI usage. Ed Newton-Rex, a former tech executive turned outspoken advocate for creators against unlicensed AI training, says he views this as YouTube “penalizing the little guy” after years of advertising the platform as a means of “leveling the playing field” between indie creators and big stars.
“It’s outrageous,” adds Newton-Rex. “And I suspect if people aren’t angry about it, they don’t know about it.”
Unlike self-represented musicians, labels and publishers negotiate individual licenses with YouTube to govern the use of their artists’ music. Whether AI training is permitted likely differs from company to company; Universal Music Group (UMG), for example, said this fall that its latest licensing renewal with YouTube “secured really important guardrails” around generative AI.
Bradfield Biggers, a transactional music lawyer at Halloran Farkas Kittila, tells Billboard that UMG and the other majors likely don’t have to worry about Google using its artists’ YouTube content for any AI training that they haven’t already agreed to, since “major-label agreements are typically highly negotiated, nuanced and comprehensive.” Many of the bigger indie labels also have solid negotiating power when it comes to licenses because of their work with Merlin.
“That may be less true for smaller labels and publishers, which often lack the same negotiating leverage, legal resources or proactive updates needed to address the rapidly evolving AI environment,” says Biggers. “In those cases, YouTube and other major platforms may see greater room to argue that existing language permits certain training uses, though the answer will ultimately turn on the specific contract language and facts at issue.”
Related
New Federal Music Law Would Let Indie Artists Team Up Against AI & Streaming: ‘Level The Playing Field’
Even for UGC music uploads, it remains to be seen whether Google’s legal argument around AI training will hold up in court. A federal judge is going to be tasked with deciding whether the YouTube terms of service do actually grant Google the broad license that it is claiming. If the answer is yes, it’s possible that other tech companies with social media arms, such as Meta, will be inspired to make a similar legal argument around training AI on UGC uploads.
Ken Anderson, a music litigator at Rimon who’s been watching the case but is not affiliated, tells Billboard that it’s wise of Google’s attorneys to turn to this novel argument about YouTube’s terms of service rather than rely on the more common fair use defense. “I think it’s very smart of them to make a contract argument,” he says. “Contracts are sacred in our country. Our capitalist society relies on them.”
That said, Anderson suspects that the indie musicians bringing this lawsuit might have an opening to argue that YouTube’s terms of service are unenforceable when it comes to AI training. That’s because under the law, courts can void a contract that is found to be an unconscionable contract of adhesion. “Adhesion” means a consumer has no other option but to sign the contract — a standard that he says is met in this case. “YouTube is one of the most necessary outlets for anybody who wants to have a career in popular music, period,” says Anderson. “End of discussion. There is no alternative.”
The “unconscionable” part of this inquiry is a closer call; in this second step, a judge must look at a contract and decide whether it’s deeply unreasonable or deceptive. Anderson says the broad language in YouTube’s terms of service could potentially meet this standard, meaning the judge could theoretically decide it cannot be enforced as to AI training. This outcome is rare, though.
“The statistical success rate [of this legal argument] is very low, but it’s definitely raised by the facts here as being a serious contender,” Anderson says.
A representative for Google did not return a request to comment for this story. In a previous statement to Billboard when Lyria 3 was rolled out in February, a Google rep said the model is mindful of copyright compliance and only trains on music that YouTube and Google have “a right to use under our terms of service, partner agreements and applicable law.”
While the court process continues to play out, musician advocates are arguing that the YouTube training controversy demonstrates how artists deserve a seat at the negotiating table with Google and other tech companies making AI music deals. MAC director Gubitz says this is a great argument for Congress to pass the Protect Working Musicians Act, a proposed federal bill that would allow indie artists to team up and negotiate with AI companies as a unit.
Harrison of A2IM says Google and other tech companies should realize that “partnering with creators is a better solution,” and “long term, it’s not good or healthy for anyone to try to win on these little technicalities.”
“The healthier path, and the one A2IM will keep advocating for,” he adds, “is platforms considering the best interests of the creative community rather than leaning on the fine print.”