Crashing Melodies — Generative AI, Copyright, and Music
Generative artificial intelligence (AI) has been at the top of most news agendas for the past several months. The release and rapid rise of services like ChatGPT and DALL-E has led to many worries about the role of AI in life, the implications for the economy and society, and concerns that AI-generated music could negatively impact the music market. With applications like Photoshop and Canva already integrating Stable Diffusion, and Sequoia Capital projecting that AI will be better than humans at most tasks like writing, drawing, and coding by 2030, these worries are not unfounded. If humans need brain challenges and creativity to flourish, what will happen if these are no longer needed? And even more damning are the predictions of possible extinction of the human race.
So many pieces are moving rapidly in music as well. CAA just signed a metaverse, computer-created artist named AngelBaby. The “Fake Drake” saga, which saw the release of a new generative-AI song titled “Heart On My Sleeve” mimicking the voices of The Weeknd and Drake, accelerated questions as to how copyright and AI will intersect, how music will be affected, whether AI-created works are infringing, and how much disruption might occur. And one of our clients, generative AI music company Boomy, received a mass takedown on Spotify. Before raising too much alarm though, remember that Auto-Tune was widely panned in the 1990s and 2000s for allowing singers to “cheat” in a recording session and fix the pitch of their vocals. The last decade in music AI began with Spotify’s acquisition of The Echo Nest and has evolved as creation tools like Splice become commonplace, and many others like Loudly, Supertone, and Splitter.ai find a niche in the market. AI’s current steps forward are exciting and frightening, but are only the natural progression of development that’s been in motion for many years.
Because recorded music’s basic protection is rooted in copyright law, many potential legal implications have emerged from these conversations and recent developments in AI music. An early 2023 ruling by the U.S. Copyright Office decreed that it would not grant registration to any parts of a song that were AI-created. This is, however, the most substantive ruling to date, and while the EU is considering broad AI regulation, no further legislation is currently proposed in the Western world to address the intersection of AI and copyright. Also, the “Fake Drake” saga likely raises more of a right of publicity claim by the artists, given that it did not contain any portions of prior works. A claim could be made that “Heart On My Sleeve” infringed not a work by Drake or The Weeknd but instead their distinct characters and voices. While this is settled law for fictional characters, there may remain a question as to whether real people can have their characters and voices infringed.
There are also questions about whether generative AI platforms need a license to train their systems on music. Japan flatly said no earlier this month, while leaders in Washington, DC have proposed rules requiring transparency about what is used to train AI. There might be an argument that the technology needs a license akin to a “server copy” under copyright law, where licenses are required in order to power technological processes. And further, if indeed AI-generated works are deemed not eligible for copyright protection, as the U.S. Copyright Office recently suggested, then AI works could be copied ad nauseum and human musicians could borrow freely from them to make their own works, perhaps incorporating their own AI hooks.
Meanwhile, despite their concerns about AI and music, labels have been actively partnering with companies like Endel. Endel is a so-called “leanback” AI music generator, meaning it helps to create meditative, sleep, and white noise sounds. UMG’s recent partnership will allow Endel to take stems from its artists’ music to create songs that change to fit various moods. One might also imagine a world where labels, who excel at marketing, decide to create and release AI artists and songs themselves, with no human behind them (thus avoiding the “costs” of musicians). It’s also possible that my kids might soon be able to dial up a custom version of a country artist covering a pop song on Alexa.
While most of the recent discussion has focused on generative AI, we must remember that there are many other potential uses of AI for music. Significantly greater efficiencies can be achieved via AI on the distribution and administration front to, for example, combat streaming fraud and reconcile metadata between parties. At Verifi, we have developed a machine learning data scoring system that assigns a reliability score to music metadata, giving our clients a “confidence level” in each piece of data. AI is also already in use to enhance content recognition of derivative works, and we are trialing work with ChatGPT to fill in text-based data gaps where possible.
It’s too early yet to tell how AI will impact music communities. It is likely that film music and other production music could be done by computers alone, and there will undoubtedly be other changes, but we should remain optimistic about the future. Every prior evolution of music has led to a greater music market than before. Will future pop stars be an evolution of Gorillaz, completely AI-generated, and backed by major labels? Will the public care? Will their works be protected by copyright? Or will AI predominantly be used as a tool for humans to create and connect with audiences?
Let’s hope a combination of legislative leadership, ethical work by those creating AI, and thoughtful business decisions prevent humans from losing their creative spirits. We believe that creators and rights holders deserve to be properly compensated for their work. And instead of doom and gloom, let’s not forget the tremendous possibility of AI, if deployed in ethical and intelligent manners.
