A group of independent musicians and songwriters sued Google on Friday (March 6), accusing the company of training its new Lyria 3 music-generation model on copyrighted recordings pulled from YouTube without permission or payment.
The lawsuit came over two weeks after Google announced the launch of Lyria 3 on February 18, describing it as the ‘most advanced’ generative AI music model yet. The model operates within Google’s Gemini chatbot app, allowing users to create 30-second tracks from text prompts or images.
Lyria 3 was developed by Google DeepMind, the tech giant’s AI research division, and is the latest iteration of its generative music technology.
In their 118-page lawsuit filed in the US District Court for the Northern District of Illinois (read here), the group of independent artists argued that “Google copied millions of copyrighted sound recordings, musical compositions, and lyric, including at least 44 million clips and 280,000 hours of music,” and used these to train Lyria 3.
The all-indie roster of plaintiffs includes New York-based singer and songwriter Sam Kogon, along with plaintiffs who also sued AI music generators Udio and Suno in October: David Woulard, indie R&B group Attack the Sound, father-and-son folk rock duo Stan and James Burjek, Chicago-based band Directrix members Berk Ergoz, Hamza Jilani, Maatkara Wilson and Arjun Singh, as well as Los Angeles-based composer Magnus Fiennes, and Atlanta-based songwriter and producer Michael Mell.
The latest lawsuit names Google as the sole defendant. The plaintiffs structured their lawsuit as a proposed class action.
“Google didn’t just have access to Plaintiffs’ music; it operated the infrastructure through which much of that music reached the world.”
While the earlier lawsuits against startups Suno and Udio targeted their music generators, the latest case against Google highlighted Google’s “structural leverage” as it doesn’t just operate a music generator. It also owns YouTube and runs Content ID, YouTube’s rights-management system that artists and labels rely on to police unauthorized use of their work.
The artists’ lawyers wrote: “Google didn’t just have access to Plaintiffs’ music; it operated the infrastructure through which much of that music reached the world. Google owns YouTube, one of the most important platforms for music discovery, distribution, and monetization, and it runs Content ID, YouTube’s rights management system.”
“Google had every opportunity to develop this product legally… It has the technical infrastructure, financial resources, and industry connections to clear rights before training.”
“Its database contains over 50 million reference files: full-length audio recordings provided by rights holders, including metadata identifying the artist, the rights owner, and the territories of ownership.”
The plaintiffs argue that gave Google a structural leverage over artists, labels, and publishers. “Google had every opportunity to develop this product legally… It has the technical infrastructure, financial resources, and industry connections to clear rights before training.”
Google launched Lyria 3 on February 18 through its Gemini app, making the model available to more than 750 million monthly active users. Lyria 3 generates up to 30-second audio clips complete with vocals and lyrics. Users can also upload photos or videos and have Gemini compose an accompanying soundtrack.
In a blog post announcing the launch, Joël Yawili, Senior Product Manager at Gemini App, and Myriam Hamed Torres, Senior Product Manager at Google DeepMind, said: “The goal of these tracks isn’t to create a musical masterpiece, but rather to give you a fun, unique way to express yourself.”
Google has not specified how Lyria 3 was trained, but said in the same blog post that it has sought to “develop this technology responsibly in collaboration with the music community” and has “been very mindful of copyright and partner agreements” in training the model.
MBW understands that to mean the training for Lyria 3 uses music that YouTube and its parent company, Google, “have the right to use” under their “terms of service, partner agreements, and applicable law”.
The plaintiffs disputed that framing, pointing to Google’s own published research in 2022, in which Google researchers described how they gathered “about 50 million internet music videos, extracting 30-second audio clips from each, and keeping roughly 44 million clips after filtering, representing nearly 370,000 hours of recorded music.”
The indie artists also cited a 2023 paper describing a separate training on 5 million clips totaling 280,000 hours of audio.
“Both papers were published in peer-reviewed venues under the names of Google researchers. Neither mentions a license. Neither discusses a consent mechanism. Neither identifies a single rights holder whose permission was sought,” according to the lawsuit.
The plaintiffs claimed that Google “knew these practices created legal risk” before it released any product to the public. In 2023, Google finished MusicLM, but initially withheld the model from public release as the company acknowledged concerns about cultural appropriation and misappropriation of creative content, the court filing said.
As MBW previously reported, MusicLM became available for the public in May 2023. The tool worked by typing in a prompt like “soulful jazz for a dinner party”. The MusicLM model then created two versions of the requested song for the person inputting the prompt. Users can then vote on which one they prefer, which Google at the time said will “help improve the AI model”.
The lawsuit also noted that Udio — which the plaintiffs (except for Kogon) previously sued for copyright infringement — was formed by four researchers at Google DeepMind: David Ding, Charlie Nash, Conor Durkan, and Yaroslav Ganin.
“Within months, all four left Google DeepMind, formed a new company, and publicly launched Udio, another AI music-generation service trained on copyrighted recordings without authorization,” the lawsuit said.
In addition to their lawsuit against Udio, the AI startup also faced a lawsuit from the RIAA on behalf of the major labels. After Udio settled with Universal Music Group and Warner Music Group, Google launched Lyria 3, the plaintiffs said.
Days after Lyria 3 launched, Google announced its acquisition of ProducerAI, the AI music creation platform formerly known as Riffusion, bringing the startup and its team into Google Labs.
The plaintiffs described ProducerAI as a standalone platform for creating, sharing, and publishing “full-length songs” with “dynamic vocals,” available on both free and paid plans.”
“Across MuLan, MusicLM, and the Lyria product line, Google has never specified whose music was used to train these systems, who consented, or who was compensated,” the plaintiffs’ legal team wrote.
The claims cover copyright infringement on sound recordings and compositions, removal of copyright management information, circumvention of technological access controls, false endorsement under the Lanham Act, and several Illinois-specific privacy and consumer protection claims, including alleged violations of the state’s Biometric Information Privacy Act for the extraction and storage of voiceprints to develop Lyria 3.
MBW has reached out to Google for comment.
Music Business Worldwide
