From AM Radio to AI Music

Generative AI has brought a number of gifts and burdens to the world, particularly in the field of intellectual property law. AI has challenged norms about what can be patented, who (or what) can create artwork, what rights of publicity celebrities have to defend themselves regarding AI imitations of their likeness, and where those rights and limits begin and end.

Among the gifts AI offers is the ability to create music in the style and voice of an artist, a tool which has allowed the remaining members of the Beatles to extract vocals from an unfinished piece by John Lennon to create a new song. With only half of the original band members still alive, AI allowed fans to once again enjoy the thrill of hearing a new song produced by their favorite band.

Paul McCartney praised AI’s capabilities, but also noted potential areas of conflict, particularly regarding the ability of AI to “rip off” musicians, making it nearly impossible for them to earn a living. The sound, voice, and style of any artist with work in the public domain can be used by AI to create similar works that listeners may mistake as an original work. While the original artist provides the style and inspiration prompting AI’s output, they receive no credit or compensation. It may seem like harmless entertainment now, but bears the potential to render musicians obsolete in the creation of their own style of music after a certain point in their career. Without adequate protection, it is not hard to imagine musicians feeling discouraged from sharing their creativity in the first place. This creates a question of how much use of AI should be allowed in the music industry.

Though this question is decidedly modern, guideposts for answering it can be found in a historic case. In Tribune Co. v. Oak Leaves Broadcasting Station, the court was faced with questions stemming from developing technology in the 1920s, and considered whether radio waves could be property. No court before this had yet considered if, how, and who could own radio waves, and when those rights are infringed. A larger radio station had been using a particular frequency for a long period of time, and a secondary local station began using a close frequency, causing users tuning into the larger station to sometimes hear the smaller station’s programming. After determining that the larger station had a property right in its radio frequency, the court had to determine whether the local station’s frequency, being not exactly the same, nevertheless should be enjoined. According to the court, the test was “whether or not there is such an interference by the defendants with the broadcasting station of the complainant.” In applying this test, the court recognized the importance of the capabilities of technology, the understanding of the tech as it stood at the time, and the physical locations where the frequency was being used. Ultimately, the court concluded that frequencies “sufficiently near to [each other] so as to cause any material interference” were too close, and also concluded that where that owner’s frequency failed to reach, there may exist an issue as to ownership.

Despite being tried nearly 100 years ago, this case demonstrates a considerate approach regarding technology with bounds yet to be defined. And although the scope of AI-generated music will depend on the specifics of various intellectual property laws, this case about radio waves offers a helpful insight. Like the rights in radio waves that Tribune recognized, many IP rights like copyright define their boundaries based on “central claiming,” in which the right to exclude emanates from a source and usually reaches as far as use and utility permit. This can be visualized as a stake in the ground, with anything close enough to the stake belonging to the owner of the property. This approach is in contrast to peripheral claiming, where boundaries are defined in anticipation of a future dispute. Central claiming takes a more ex post approach, deciding boundaries as conflicts arise, instead of defining the boundaries as to prevent such conflict.

Thus, the “interference” logic of Tribune can help to guide future IP decisions on AI-generated music, even if the case does not apply as a legal matter. Consider, for example, a song recording that uses an AI-generated voice simulating Paul McCartney. Regarding “material interference,” McCartney would argue that listeners would naturally associate the AI rendition with the Beatles, and may not even be able to distinguish the difference. Listeners would enjoy music of “the Beatles,” or rather something similar enough, without due credit or compensation to the actual artists. Further, to an extent, the Beatles would lose control over their own sound and reputation. The AI user, on the other hand, would argue no material interference because listeners would be able to distinguish the difference, and would not mistake an AI creation for an original Beatles song. Tribune also suggests that the state of the technology is relevant, which favors McCartney because AI is a relatively new and rapidly developing tool, which may suggest that users would less likely be able to keep up and distinguish an original from a fake. And regarding physical locality, one could argue that AI creations popularized in the US should not interfere with an artist in the UK, however the internet continues to make locality less relevant in this regard considering an AI creation can be shared with overseas listeners instantaneously. Based on these arguments, McCartney would be more likely to win because interference would be highly likely.

These approaches are also applicable to users of AI and developers alike. An interference-based approach could help companies developing AI to expand and progress their programs more rapidly, creating a field that encourages consistent growth. When AI reaches the point of creating interference in the mind of consumers, causing them confusion as to the source of the creation, the boundary of fair use of AI has been reached. While this may cause greater legal conflict, it would allow for AI to define the boundaries as it crosses them, instead of stifling them on the front end, ex ante.

Stifling AI would limit the benefit it could provide to its diverse range of users. The Beatles using AI to create a new song half a century after they disbanded is just one example. While it may seem prudent to preemptively define the bounds of what AI can and cannot do, such as placing restrictions to guard against deeps fakes and misuse of voice imitation by AI, the positives that AI provides offers exponentially more benefits with increased freedom. Taking the potential benefits into consideration, it may well be in society’s best interest to allow AI’s momentum to continue. When issues arise, the Court may look to predecessors 100 years before in developing new law for the regulation of AI.

Allison Guenther is a 1L at the American University Washington College of Law and a Junior Staffer on the Business Law Review.

Joseph Rios is a 1L at the American University Washington College of Law and a Junior Staffer on the Journal of Gender, Social Policy, and the Law and a Senior Staffer on the Intellectual Property Brief.

Image: United Press International, photographer unknown, The Beatles arrive at JFK Airport.