AI-Generated Music: Copyright’s next frontier

In the early years of the twenty-first century, Napster changed the music industry.[1] Soon enough, Artificial intelligence will also dramatically change the music industry, as is already the case across all fields. With the evolution of AI, laws will need to adapt to safeguard the rights of industry stakeholders.

What was once considered far from reality was made quite apparent by a track downloaded by TikTok user Ghostwriter. The track called “Heart on my sleeve” was created using AI technology to replicate the voices of Drake and The Weeknd, ultimately churning out a track that raked in millions of views[2]. A few days later, Universal Music Group (which has dealt with both Drake and the Weeknd) sent letters to all streaming platforms, including YouTube, Apple Music, and Spotify, demanding the removal of all songs replicating artists from their platforms, alleging copyright infringement.[3]

The issues faced by the music industry concerning the use of AI are those all IP experts are facing concerning the topic: Since copyright protects works made by a “human spirit,” who owns the copyright to a song made by an AI? Some have answered that the AI’s developer should own the copyright, while others argue that the user, who used AI as a mere tool (akin to a pencil), should own the rights. However, in our case, we find even more stakeholders with vested interests in rights ownership; what about the artist whose voice has been cloned to produce the song? How about the artist’s label if he is signed to anyone who owns the recordings?

AI is transforming how Music is being created; AI tools are revolutionizing music creation in diverse ways. AIVA[4] is a creative assistant that enables musicians to generate compelling themes leveraging AI swiftly​​. At the same time, Amper[5] allows content creators to control the length, structure, and mood of their Music, ensuring that creations match their narrative​​​. Soundful's[6] platform rapidly enables creators to generate unique, studio-quality Music, offering a wide range of genres and customization options​4​. Simultaneously, ChatGPT assists in the songwriting process, generating lyrics or completing a song based on a given title or theme, showcasing the potential of AI-assisted music creation​.[7] These tools collectively exemplify AI's transformative role in democratizing and accelerating music creation.

AI is thus slowly but surely gaining prominence in music creation, but these creations remain without a definite owner considering the silence of legislation on the subject.  However, in a promising move, the U.S. Copyright Office recently issued new guidance regarding AI-generated works clarifying that artistic works created with the help of artificial intelligence are copyright eligible depending on whether AI's contributions are "the result of mechanical reproduction," such as in response to text prompts or if they reflect the author's "own mental conception."

According to the Office, the answer depends on how the AI tool operates and how it was used to create the final work... The office reiterated that copyright protection depends on the amount of human creativity involved, and the most popular AI systems likely do not create copyrightable work. Creative modifications and arrangements of AI-created work can still be copyrighted, and applicants must disclose when their work includes AI-created material. The Office understands that users do not exercise ultimate creative control over how such AI systems interpret prompts and generate material. These prompts function more like instructions to a commissioned artist​. [8]

As such, especially considering the lack of answers provided by legislation or statute, the US copyright office provided the music industry with some guidelines to help it navigate the tides of AI. Going back to the basics, it holds that human creativity and involvement are paramount in making work copyright eligible. One must then look at the producer’s involvement; did he create the music and write the lyrics, or were they AI-generated and the result of a couple of prompts? Answering the question of the artist’s involvement would help solve the issue for now before legislators undertake further developments to set clear and well-defined rules.

However, one question remains prevalent: What about voice impersonation? Vocal impersonations have been around before the emergence of AI, and courts have already dealt with this issue, most notably in Midler v. Ford Motor Co. (1988).[9]

In this case, Bette Midler sued Ford and its advertising agency for imitating her voice in a commercial for the Ford Lincoln Mercury after she declined to participate. A former backup singer was used to impersonate Midler's voice, and many people commented that the commercial sounded exactly like Midler's record of "Do You Want to Dance." Ford and the advertising agency had obtained a copyright license to use the song but did not use Midler's name or picture in the commercial. The U.S. Court of Appeals for the Ninth Circuit evaluated Midler's claim under the Federal copyright law and state tort law. Under Federal copyright law, the Court found that imitating a recorded performance, even if deliberate, does not constitute a copyright infringement. However, under California law, the Court found that a person's voice is as distinctive and personal as their face and that impersonating someone's voice is equivalent to taking their identity. The Court held that when a widely-known singer's voice is deliberately replicated for commercial use, the imitators are liable under tort law. Midler won $400,000 in damages on her tort claim. Thus, if AI is used to mimic a person's voice without their consent, similar to the impersonation of Midler's voice in the Ford commercial, it could be an infringement of the individual's right to publicity, depending on the jurisdiction.

Therefore, universal’s claim cannot be based on voice impersonation as held in Midler vs. Ford, whereby voice is not considered protected by copyright. What was the request of removal based on then?

According to UMG, copyright infringement is rampant during the AI's training, whereby the developers are illegally feeding copyrighted music to the software to learn how to mimic artists' voices and music. UMG's aims to prevent AI from becoming a potentially existential threat to the music industry, particularly AI so advanced that it can recreate melodies and artists' voices in seconds. UMG sees this as violating their agreements and copyright laws and has asked streamers to remove AI-generated songs. A Universal Music Group (UMG) spokesperson conveyed to CNN that the company views it as a moral and commercial responsibility to their artists to hinder unauthorized music usage.[10] The effectiveness of UMG's move is debatable, as AI services may still access copyrighted material. There are no regulations on what AI can and cannot train on. This issue has raised significant questions about copyright and the definition of an artist in the age of AI. It also brings to light the fundamental legal and ethical responsibilities of platforms to prevent their services from being used in ways that harm artists. This is especially concerning considering the non-consensual use of an artist's voice that may sometimes be used for harmful purposes, such as the AI-generated Drake cover of "Munch," over which the artists clearly expressed discontent. [11]

In a progressive move, Grimes, the acclaimed Canadian artist, has presented an innovative solution to the ongoing debate about using generative AI in music creation. Recognizing the burgeoning trend of AI-generated music, she has extended an invitation to musicians worldwide to create new songs incorporating her voice via AI technology. This groundbreaking initiative promises to split 50% of royalties on any successful AI-composed track that features her vocals, echoing the same terms she offers for traditional artist collaborations.[12] Grimes even went further, launching a platform that helps users make AI songs using her voice.[13]

AI is no longer science fiction; it is a reality. Eventually, the music industry will have to adapt to technological developments through the development of laws or innovative solutions by critical stakeholders. As AI technology develops, we will likely see even more creative and novel uses of AI in the music industry. This could lead to new challenges for copyright law, as it will become increasingly difficult to determine who owns the copyright to AI-generated works as much as we do not have such facts. Furthermore, a fundamental question has to be answered: How can a balance be struck between innovation and the protection of artist rights? Should requests such as the one by UMG be seen as oppressive and stifling a natural evolution in the sphere of music creation? Unfortunately, we do not have the answers; we can only hope that legislators and stakeholders make the right decisions to protect the rights of all artists while enabling creators to use AI technology to advance their craft.