Back in 2013, a controversy shook the music industry when Robin Thicke and Pharrell Williams allegedly plagiarized Marvin Gaye’s song “Give it Up” in their pop (and sexist) hit, “Blurred Lines.” The lawsuit began when Gaye’s children sued Thicke and Williams for stealing their father’s song. Gaye’s offspring were ultimately awarded millions of dollars due to copyright infringements by Thicke and Williams. After reading about this particular dispute, I was puzzled just how exactly one can determine if a song is plagiarized, especially since artists have been copying each other’s music for centuries (think Beethoven versus Mozart)! Subsequently, I ran a quick Google search on this issue, where I learned about a new Artificial Intelligence (AI) technology, created and used by streaming services in order to detect plagiarism on their platforms. As I began my initial search, the motivations for creating these sorts of fraud detection technologies appeared to be a little fuzzy. I was unable to determine who the detection system would protect. Was the intention to safeguard the streaming service, artist or original content? In addition, it appeared as though the technology was a positive invention, much like the Turnitin feature students use on Canvas. I began to wonder just how this AI fraud detection technology is disrupting the arts world? To understand this vast disruption, I had to start at the beginning with the foundations of AI technology, copyright law in the United States and the role music plays in peculiar situation.
Simply put, AI is a technology that utilizes computer science together with data to create solutions to problems. The idea of AI technology has been around since 1950 with the British mathematician, Alan Turing. However, the computers were not sophisticated enough to build a machine where they could be programmed to complete tasks and commands. AI largely advanced in the 1970s and have only improved since computers’ capabilities have expanded. Today, AI is used for a multitude of projects and problem-solving techniques. AI has greatly shaped the music world as computer scientists create new technology to help compose music, play music, and even detect plagiarism.
US Copyright Law
Since the creation of the United States Constitution in 1787, copyright law has been weaved into American business entities and society. In Article I, Section 8, the Constitution states, “Congress shall have the Power…to promote the progress of science and useful arts, by securing for limited times to authors and inventors the exclusive right to their respective writings and discoveries.” Already at the establishment of the United States, Congress allowed protections for those who produced original works, in addition to having the ability to receive royalties for their work(s). Since the ratification of the Constitution, US Copyright law has evolved. In 1909, Congress passed the first Copyright Act of 1909, which was signed into law by President Teddy Roosevelt. This law “granted protection to works published with a valid copyright notice affixed on copies” in addition to granting protection for unpublished works. The Copyright Act was revised in 1976, which is the current a copyright law in effect today.
The Copyright Acts extends to many forms of property, including music. A music copyright gives the owner the “right to make and sell copies, distribute those copies, make new works, and publicly perform the work.” The copyright law also enables music owners the right to earn royalties from their work.
There are two components in music copyright: composition and sound. When a song is recorded, there is the composition. This includes whatever notes, rhythm, melodies, or chord arrangement/progression used. The other component is the sound recording which encompasses the sounds of the song like lyrics, which are incorporated into the song recording. The sound piece is typically created by the artists and producer, who also hold the rights to this piece of the song. The interesting aspect of a musical copyright is the protection occurs as soon as the original piece of music is “fixed,” or the music is recorded in an audio file or even notated on sheet music.
To determine whether an artist stole or borrowed someone else’s work, involves a two-tiered test. The first part involves something called access, in which there is evidence that that artist heard or “be presumed to have heard” the original piece of music before writing their song. The second part involves something called the “substantial similarity,” where “the average listener can tell that one song has been copied from the other.” This test is quite problematic as someone needs to catch an artist in the act of listening to a copyrighted song or claiming that a song has been stolen based on just listening to it. Just like visual art, music tastes differ from person to person. People do not hear things the same way as others do. This can lead to a case of a he-said-she-said situation resulting artists denying that they ever plagiarized.
However, with new technology, the two-tiered plagiarism test for music may no longer be necessary. Technology has evolved immensely to the point where streaming platforms are developing their own technology and devices to detect plagiarism in copyrighted material. As a Spotify user, I heard of their efforts to curb plagiarism through a new technique that they developed. Because I use Spotify so frequently, I was interested in this new technology.
Spotify’s New Invention
In 2020, Spotify filed for a European patent for a new technological invention that computer scientists had created. Using AI technology, the audio streaming software conglomerate developed a system to detect fraud and plagiarism within uploaded content to their site.
With this new technology, Spotify set out to form a solution to the common problem of people stealing copyrighted music. However, Spotify’s detection machine uses a unique method for catching fraud, involving music lead sheets.
A lead sheet is simply the components of a song which includes, melody, harmony, chords and lyrics. After the completion of writing a song, the lead sheet is fed through this detector system and compared to the countless other lead sheets in the Spotify library. The technology will then inform the artist of any similarities. Not only does this technology prevent fraud, but it occurs almost in real time, so artists are able to revise their songs right away, in order to avoid any potential lawsuits.
As a musician, I was curious to see if one of my previous compositions would be detected as having any elements of plagiarism. So, I decided to attempt to use Spotify’s invention. Granted, my old composition is a piano piece, so the Spotify library would need to identify piano music lead sheets to compare with my own composition. Because I wrote this piece in the key of A minor with basic chord progressions, I thought it would be quite simple to find elements of plagiarism within my piece.
Minutes into my quest, I quickly ran into a problem. In order to run a piece of music through the fraud detector, an artist must upload their song to Spotify as a piece of music available to everyone. I could upload my composition; however, it would only be in my personal Spotify library of my personal uploads. The only way to make my music public, was to go through a distributer. Spotify has a list of their “preferred distributers,” where the artist pays a monthly fee to the distributer who will upload their songs onto Spotify. The artist will then be able to track listenership, streams, and test the similarity of their song(s). Since I was unwilling to pay a service to upload my composition, I was unable to complete my experiment.
After going through the motions of trying to upload a piece of music, I realized Spotify’s rationale and motivation behind this technology and it all boils down to money. Spotify does not want to be sued and found responsible for publishing potentially risky music. As I previously stated, music copyright tests were primarily auditory and needed substantial proof to bring on a lawsuit. Spotify’s technology acts as a waiver for artists to sign, declaring no responsibility to any published music.
While the technology world looked in awe at Spotify’s invention, the ramifications and disruptions to musical artists are vast. To begin with, there are only 12 notes in the Western musical scale. In addition, popular music largely uses the same chord progressions, which typically involve patterns of four chords in a specific key. This poses many challenges to artists as they write songs because there are only so many notes and chords available. Musical keys and chord progressions are not copyrighted so artists are bound to have used a previous chord progression and notes from other songs. This places musicians in an unfair predicament as their creativity is stunted because they must worry about potential copyright issues as they pursue their craft.
Another issue regarding Spotify’s copyright machine is the livelihood of artists. This technology does not benefit the small, independent musician. As I found in my experiment, in order to publish music on Spotify and gain royalties, artists must pay a distributer fee. These fees vary, depending on the distributer, which could range as low as $10 for a single release to over $100. Additionally, Spotify will pay $0.0033 per stream, or “250 streams to earn a dollar.”
With the use of the plagiarism detection machine, if an artist’s work is found to be closely related to another song, the artist will need to revise the song until it is deemed original. This process then becomes lengthy as the revision process is time taken away from an artist gaining streams and earning royalties. At the same time, the artist is still paying a distributer to upload the songs to Spotify. This entire procedure is quite expensive and unless the artist is already wealthy, this lifestyle is unsustainable. The notion of a struggling artist only continues as Spotify wishes to avoid costly lawsuits, at the expense of the artists, in order to stream music to the world.
Bagley, Dave. “AI Generated Music: Who Gets the Royalties?” VIstex (blog), January 11, 2021.
Dorn, Lori. “Google Introduces an AI Search Feature That Identifies Songs by Whistling,
Gray, Joanne E, and Nicolas P Suzor. “Playing with Machines: Using Machine Learning to Understand Automated Copyright Enforcement at Scale.” Sage Journals, Big Data & Society, April 28, 2020.
Gershgorn, Dave. “Is Spotify’s Newly Patented A.I. Plagiarism Detector a Data Collection Scheme?” OneZero, December 7, 2020.
Heffler, Jason. “Spotify Has Invented AI Technology for Songwriters to Detect Plagiarism.” Emd.Com, December 2, 2020. https://edm.com/gear-tech/spotify-invented-ai-technology-detect-songwriter-plagiarism.
Jacob, Ennica. “How Much Does Spotify Pay per Stream? What You’ll Earn per Song, and How to Get Paid More for Your Music.” Business Insider, https://www.businessinsider.com/how-much-does-spotify-pay-per-stream. Accessed 28 Feb. 2022.
McGlynn, Declan. “AI Futures: How Artificial Intelligence Will Change Music,” October 5, 2021. https://djmag.com/longreads/ai-futures-how-artificial-intelligence-will-change-music.
Pastukhov, Dmitry. “6 Basics of Music Copyright Law: What It Protects and How to Copyright a Song.” Soundcharts (blog), February 10, 2020.
Provider Directory – Spotify for Artists. https://artists.spotify.com/providers. Accessed 28 Feb. 2022.
Stassen, Murray. “Spotify Just Invented AI Technology That Will Police Songwriter Plagiarism.” Music Business Worldwide, December 1, 2020.
Sturm, Bob, Maria Iglesias, Oded Ben-Tal, Marius Miron, and Emilia Gomez. “Artificial Intelligence and Music: Open Questions of Copyright Law and Engineering Praxis.” MDPI, September 6, 2019.
Tada, Yuriko. “The Internet and Musical Copyright.” Harvard Law School, 1998. https://cyber.harvard.edu/fallsem98/final_papers/Tada.html.
“3 Ways Spotify’s Plagiarism Tool Is Different.” Plagiarism Today, 7 Dec. 2020, https://www.plagiarismtoday.com/2020/12/07/3-ways-spotifys-plagiarism-tool-is-different/.
Timeline 1900 – 1950 | U.S. Copyright Office. https://www.copyright.gov/timeline/timeline_1900-1950.html. Accessed 28 Feb. 2022.
Timeline | U.S. Copyright Office. https://www.copyright.gov/timeline/. Accessed 28 Feb. 2022.
Titlow, John Paul. “YouTube Is Using AI to Police Copyright- to the Tune of $2 Billion in Payouts,” July 13, 2016. https://www.fastcompany.com/4013603/youtube-is-using-ai-to-police-copyright-to-the-tune-of-2-billion-in-payouts.
TRT World. “Fun or Fraud? The Rise of Deepfake Voice Technology Can Make the Dead Sing,” July 14, 2021. https://www.trtworld.com/magazine/fun-or-fraud-the-rise-of-deepfake-voice-technology-can-make-the-dead-sing-48373.
Vk, Anirudh. “How YouTube’s ML Algorithm Earned Billions for Music Producers by Building Fingerprints for Songs.” Analytics India Magazine, March 26, 2019. https://analyticsindiamag.com/youtubes-ml-algorithm-earned-billions-music-producers-building-fingerprints-songs/.
What Constitutes Music Plagiarism? https://lawyerdrummer.com/2017/03/music-plagiarism-2/. Accessed 28 Feb. 2022.
What Is Artificial Intelligence (AI)? https://www.ibm.com/cloud/learn/what-is-artificial-intelligence. Accessed 28 Feb. 2022.
What Musicians Should Know about Copyright | U.S. Copyright Office. https://www.copyright.gov/engage/musicians/. Accessed 28 Feb. 2022.