Congress Should Fund the Creation of a Similarity Checker for Music
A Los Angeles jury found in 2015 that “Blurred Lines”—Pharrell Williams’ hit song with Robin Thicke—had illegally infringed on the copyright of Marvin Gaye’s “Got to Give It Up.” This was a landmark ruling because prior cases had only extended copyright protection to lyrics and melody. In Williams v. Gaye the court allowed several other elements, such as harmony, pitch, rhythm, timbre, and tempo, to be considered as well. As a dissenting judge on the appellate court later wrote, this decision allows someone to “copyright a musical style” and “establishes a dangerous precedent that strikes a devastating blow to future musicians and composers everywhere.” As a result, it is now harder for artists and record labels to determine where permissible influence and interpolation become impermissible appropriation and plagiarism.
Whether a musical work crosses the line for copyright infringement was already blurry (no pun intended). Judges or juries, depending on the specific legal procedure and jurisdiction, evaluate the works in question to determine if there is sufficient similarity between them to warrant a finding of infringement. The test can involve both an objective comparison of the works’ elements and a subjective assessment of whether an ordinary observer would recognize the similarities between them. But the test to determine infringement is different in court circuits across the country. Some courts do both tests, some courts only do the objective comparison, and other courts do some variation. Congress has never established a standard for this test and the Supreme Court has not provided clarity in this aspect of the law.
Congress should make things more consistent, accurate, and fair by directing and funding the Copyright Office to launch a competition for the private sector to come up with an AI-enabled tool to compare how similar a musical composition or recording is to existing copyright-protected works. This service should be modeled after the popular Turnitin online service that educators use to check how similar their students’ written submissions are to existing written works. Importantly, these tools do not claim to identify plagiarism, instead, they simply flag similarity and provide educators with the information they need to make a judgment. They can work to check similarity regardless of whether a work was written exclusively by a human or with the help of generative AI tools like ChatGPT.
An AI similarity checker for music could help identify how similar new songs are to existing songs made by humans as well as those made with the help of new AI music creator tools such as Google’s Lyria, and empower judges and juries to decide when plagiarism has occurred.
There are lots of reasons why this type of tool would be beneficial. For one, such tools could reduce the human bias inherent in music-related judgment calls. Regardless of whether a court uses the objective test or subjective test, they rely on humans to decide how similar two pieces of music are. But human testimony tends to be inconsistent because plaintiffs and dependents each hire their own musical experts who focus on either the differences or similarities of songs depending on which side they are on, which only confuses courts according to law experts Angelyn Gemmen and Sean Tu at West Virginia law school. Tu argues AI can act as a substitute and “courts would no longer have to rely on a battle of the experts or the use of lay observers to determine if a work is substantially similar to another work.” Using AI would also enable consistent analysis and more impartiality. The algorithms used and the datasets it trains on could be agreed upon by an industry standard.
Perhaps most crucially, creating such a tool would give artists the ability ex ante to anticipate whether their work likely infringes on any existing copyrighted works. Having a portal where artists can check their work before they release it gives them an opportunity to identify areas of risk and change the parts that are flagged as potentially infringing. This system could increase efficiency because producers and record labels could use the tool to identify liability risk and make more targeted decisions about when to purchase expensive liability insurance in cases and when to make artists change parts of their songs, leading to more expressive works.
Congress has an opportunity to help unblur the property lines of music. It should act now to pave the way for a more equitable, efficient, and innovative music industry.
