Judith Finell

ELVIS Law and Fake Drake

States have long had laws restricting the use of a celebrities’ “name, image and likeness” for commercial purposes such as advertising. The issue is gaining momentum because of AI generated “deep fakes” such as the notorious “Fake Drake” song that dropped on streaming platforms recently. To address this risk, Tennessee has now enacted the Ensuring Likeness, Voice and Image Security Act (ELVIS Act), which bars the use of a person’s identity as a singer “regardless of whether the sound contains the actual voice or a simulation of the voice of the individual.”

The ELVIS Act holds technology providers responsible and liable if their purpose “is the production of a particular, identifiable individual’s photograph, voice, or likeness, with knowledge that distributing, transmitting, or otherwise making available the photograph, voice, or likeness was not authorized by the individual.” The Act went into effect on July 1, 2024.

The Finell Hierarchy of Elements in Assessing Substantial Similarities

In evaluating music enmeshed in plagiarism disputes, the foremost questions a musicologist addresses are:

Are these works alike?
If so, how are they alike?
Are their underlying compositional features alike?

In my own musicological education and its application to copyright infringement disputes, I have developed and relied upon a hierarchy of musical features in assessing the degree of similarity between musical compositions. Most often, the elements at the top of the hierarchy carry greater weight than the lower ones.

Hierarchy of Elements in Assessing Substantial Similarity

Melody – pitch
Melody – rhythm (duration)
Harmony
Lyrics (if applicable)
Metric placement
Relative proportion of similar significant elements
Structure

Not all features are equally significant to a musical work, and some similarities are more fundamental to the expression of the works than others. There are also often overlaps between the hierarchical features. For example, metric placement describes the position of pitches and rhythms within the pattern of stressed and unstressed beats that propel a melody forward. Two songs can have a similar series of pitches and rhythms that do not align due to differing positions and stresses. The impact of stresses and position on musical expression is analogous to the English language, in which changing an emphasized syllable impacts the meaning of a word. For example, compare “inVALid” (adjective) to “INvalid” (noun), and “atTRIbute” (verb) to “ATtribute” (noun). The same sometimes applies to musical comparisons involving metric placement.

At times, the absence of one of these primary musical traits elevates the importance of items lower in the hierarchy. For example, many rap works contain rhythmic spoken lyrics with no melodic pitch. This means that the rhythm and rhyme scheme and underlying musical backdrop (referred to as the “bed”) become paramount in the comparison process over melodic elements. An extreme example of melodic minimalism are one-note songs, in which a single note is sung repeatedly with lyrics listing information. In opera, this genre is referred to as “catalogue arias,” and in musical theater, as “patter” songs. Examples include significant portions of “Johnny One Note” (Rodgers and Hammerstein), “One Note Samba” (Antônio Carlos Jobim), “I Am the Very Model of a Modern Major General” (Gilbert and Sullivan), and “(Not) Getting Married Today” (Stephen Sondheim). In the case of one-note melodies, pitch becomes secondary to rhythm, harmony, and lyrics.

Does AI Democratize Musical Creation?

AI is possibly going to become a virtual collaborator with human composers. Music is a great candidate for AI, because “it’s math,” as many say, meaning it’s highly systematized and fully measurable in time, volume, tempo, frequency, and pitch. As such, it’s a prime database for machine learning. AI extrapolates patterns and traits from previous musical works to anticipate and predict options for the human composer. It is similar to word prompts on our smartphones as we type texts. There are serious copyright issues with the training of machines on copyright protected music.

We are as musicologists at the intersection of music, technology and law, applying our musical analysis training to the legal protection of musical creativity. How did Beethoven, who was deaf and lived in isolation, come up with his Fifth Symphony? What else might Beethoven have composed today if he had had the assistance of AI to overcome his barriers? Imagine what would have happened if the gifts of technology had allowed Beethoven to hear and actually feel the vibrations of his own music? Instead of waiting for an orchestra to learn and play his symphony, he could have used AI to play it back to him with a virtual orchestra.

And it’s not just about the musical geniuses. AI now has the ability to democratize the creation of music beyond elite professional musicians. Average people, simply motivated by musical passion and potential, can create music with AI.

Music Composition Collaboration with Humans and Machines

Every new technology has challenged the music industry of its time with predictions that it will cause all creativity to stop. Still, the march forward continued, and today with AI we are facing a new technological earthquake that will likely bring the opposing sides to the legislative halls and courtrooms once again.

Of all the art forms, why is music so often on the front lines? Partly, because the battle lines are so clearly defined. The music industry has enforcers with the Recording Industry Association, the National Music Publishers Association, and the Performing Rights Organizations (ASCAP, BMI) ready to fight the battle, and well armed to do so. Music is the art form that has carved out 6 specific rights to be licensed exclusively from the rights holders and creators: reproduction, distribution, public performance, synchronization (music combined with visual and moving images), creation of derivative works, and public display (lyrics) of the works. Music creators and their representatives at record labels and publishers staunchly guard these rights.

The US Copyright Office, itself, has grappled with these issues as a constitutional right since the beginning of the USA. The tech industry is on both sides of the issue. They are both creators and users of original content, with much to gain and lose from the legal outcomes of this battle.

Music is a perfect candidate for machine learning, namely AI generated music. Most musical features of a given work are highly codifiable and quantifiable, making it fertile ground from which AI can recognize patterns, create variations, and build new compositions from the traits of previous examples. Often this is done using copyright-protected music without permission, credit, or payment to the original creators of the content. A showdown is inevitable.

The technical elements of music that are protected by copyright and used by AI as the learning mechanism, as well as the robust music protection system controlling its use are heading for a face-off. The primary musical elements involved include the very essence of musical works, namely pitch sequences, rhythms, lyrics, chord progressions, along with performance features, such as timbre (the character of a particular artist’s singing voice or instrumental performance style), recording studio and arrangement features, and more.

Are the existing legal criteria for determining substantial similarity and copyright infringement still applicable to the musical creation process with AI? The roles of expert testimony, jurors, lay listeners and more will be examined as it applies to AI generated music.