A group of independent artists are suing Google over its Lyria music AI model, which they claim has been trained on tens of millions of recordings without the necessary licences from the music industry.
In a filing with the Illinois courts, the musicians acknowledge that their legal action joins a long list of copyright lawsuits filed by creators and rightsholders against generative AI companies, some of which also relate to music AI. However, they insist, “Google’s conduct is different”.
That’s because Google, more than anyone, possesses the knowledge and infrastructure to do this the right way, say the musicians, but instead it infringed copyright just like all the other AI companies. “Google had every opportunity to develop this product legally”, says the new lawsuit. After all, Google “owns YouTube and runs Content ID”.
And it “has long-standing relationships with major labels and distributors”, not to mention “the technical infrastructure, financial resources, and industry connections to clear rights before training”.
But Google chose to not obtain licences before training its music AI model, the lawsuit concludes, “not because licensing was impossible” but because copying without licences “was faster and cheaper”.
YouTube itself has been more cautious when talking about music and AI than start-ups like Udio and Suno. When it launched a Music AI Incubator in 2023 it published a set of principles for developing music AI tools, including that it would “embrace AI responsibly together with our music partners”. Universal Music was also announced as a specific partner on the AI Incubator project.
Later that year the video platform piloted an AI voice clone tool called Dream Track in partnership with Universal Music and Warner Music, which allowed a small group of YouTubers to create music that featured AI-generated vocals in the style of Charlie Puth, John Legend, Sia, T-Pain, Demi Lovato, Troye Sivan, Charli XCX, Alec Benjamin and Papoose.
However, Google’s separate DeepMind AI division, which developed Lyria - which in turn powered Dream Track - has been a little less cautious. Yes it has occasionally discussed copyright matters in research papers and has had some talks with the music industry. But, according to reports, it has also trained its AI models on recordings and songs without specific licensing deals in place, just like the AI start-ups.
And when it launched Lyria 3 last month, an official announcement said simply that Google was “very mindful of copyright and partner agreements”. But, as the new lawsuit points out, “it has not named any agreement” and “it has not specified a single licence covering training data”.
And even if Google did get the necessary permission from the bigger record labels it has direct relationships with, what about all the independent artists whose music is believed to be within the massive training dataset used by Lyria?
Presumably Google, just like the AI start-ups, will ultimately claim that AI training is fair use under American copyright law, meaning no permission is required from creators or rightsholders. That argument is currently at the centre of dozens of lawsuits and there remains much legal uncertainty over whether or not AI training is or is not fair use.
In the context of this lawsuit, Google would presumably like a legal precedent to be set that clearly states the kind training undertaken by DeepMind is fair use. Though, ironically, if it turns out AI training is not fair use and all AI companies are forced to secure licensing deals from copyright owners, that arguably gives Google a big advantage over the start-ups.
Because, as this lawsuit points out, it has existing licensing deals and relationships with the music industry, as well as one of the best rights management platforms in the business, and a massive existing audience with which to test any music AI tools.
Though in the short-term, Google will be hoping this is a case where fair use - or any other useful legal arguments - do win the day. The musicians in this litigation are pushing for class action status, meaning the lawsuit could represent all independent artists, which could result in Google having to pay out mega-damages if a court found it liable for copyright infringement.
