Music, AI basics and how the Beatles are as human as ever

Artificial intelligence generally elicits excitement and trepidation. To better understand the legal implications of music and AI, it can be helpful to discuss 4 main categories: (1) Ingestion and training of AI systems; (2) Generative output that AI produces; (3) Right of publicity and the use of an individual’s voice, name, and likeness within AI systems; and (4) AI as a tool to assist in creation. Within each category, there are different legal concerns to analyze that can provide a basic framework of how AI and music interrelate.

First is training and ingestion. Training AI systems requires the ingestion of content to train such systems for their particular function. This ingestion and training process generally requires making a copy of material that may be copyrighted. In music, this is primarily the underlying musical compositions and the sound recordings that embody such compositions, however it can also include music videos and artwork, as well as right of publicity (as discussed below). The basic analysis to consider when training AI systems is whether the use of material is copyright infringement or fair use. The right “to reproduce the copyrighted work in copies or phonorecords” is an exclusive right held by copyright owners under 17 U.S.C. Section 106(1). Absent a defense, such as fair use, any such copying would in general be an infringement. Fair use is a balancing act codified in 17 U.S.C. Section 107 and recently discussed by the Supreme Court in Andy Warhol Found. for the Visual Arts, Inc. v. Goldsmith, 143 S. Ct. 1258 (2023). It’s important to note that these sources only deal with fair use in the U.S. Other countries have different frameworks for fair use and potential AI exceptions. Another issue with training and ingestion is the removal of content management systems. Metadata that identifies copyrighted material may be removed when AI systems ingest such material, which could be a violation of 17 U.S.C. Section 1202(b). There are a number of ongoing cases in this area, but nothing is settled yet. In the interim, understanding the copyright framework can help determine whether training of AI systems is infringement or not. If it is infringement, there’s still the option to license and pay for the rights from copyright owners. Many companies train AI with content they already own, which alleviates the concern of infringement and fair use. However, the issue still remains if other content is used and whether use is infringing or fair, as well as whether it removes any content management systems.

Second is AI output. The U.S. Copyright Office has provided clear guidance that human authorship is required in order for a work to be copyrightable in recent cases such as Thaler v. Perlmutter, Case No. 1:22–cv–01564 (D.D.C.) and the Théâtre D’opéra Spatial Copyright Office Review Board Decision. When an AI system outputs content that was prompted by the user, the output is not copyrightable because there’s no human authorship. Even if the output included copyrightable material that’s not purely created by AI, there’s still the potential that the output would violate a copyright owner’s exclusive distribution right and right to create derivative works under 17 U.S.C. Section 106. Arguably, the prompt itself could be copyrightable as a textual or literary work, but that’s different then the output. In the U.S., the output is what needs human authorship to have a copyrightable right in a work that is expressed in a tangible medium pursuant to 17 U.S.C. Section 102(a).

Third is the right of publicity. Under California Civil Code Section 3344, you need someone’s permission to use their name, voice, likeness, and image. A well-known case concerning the right of publicity is Midler v. Ford Motor Co., 849 F.2d 460 (1988), which held that impersonating someone’s voice for commercial purposes without their permission is a violation of their right of publicity. This right also includes the right to consent to and profit off of your own name and likeness. Part of the AI landscape utilizes others’ voices and likenesses. Whether this is a violation of the right of publicity is currently determined state by state; however, there is also currently a proposed Federal act called the NO FAKES Act that would protect unauthorized use of an individual’s voice and likeness from generative AI. There could also be implications at a Federal level outside of state right of publicity issues, such as if use of someone’s right of publicity creates a false designation of origin, then trademark issues could arise.

The last category discussed here is AI as a tool to assist in creation. As an analogy, think of Microsoft Word: When writing a poem or essay, writers use Microsoft Word as a tool, but there’s still a human author. The same can be true in music. The Beatles recently released a song titled “Now and Then” which used AI to separate John Lennon’s vocal stems from piano so they could mix it. AI didn’t output anything expressive that would replace the human authorship requirement for copyright ownership. Instead, AI was used as a tool to separate tracks. When AI is used as a tool, there’s still a human utilizing the tool to create an expressive work. However if AI is used as more than a tool, there’s currently a duty to disclose what was AI-generated and what was human-generated when registering with the U.S. Copyright Office to ensure claims are sufficiently original.

Note that there’s always the option to voluntarily enter into an agreement with creators and copyright owners. This allows for proper attribution, compensation, and a level of ethics in respecting both creators and innovation. There’s a balance to be struck between protecting authors under the Constitution and encouraging free expression. We’re in a new age with laws not yet clearly set. The above provides the basic framework for navigating how AI touches the music landscape in today’s day and age. We’ll see how cases, laws, and technology develop moving forward.

DISCLAIMER: This information is for informational purposes only and should not be taken as legal advice for any individual case or situation; and please do not consider this information to be a substitute for obtaining legal advice from a qualified attorney licensed in the jurisdiction(s) relevant to your matter.