Recipe for Success?: An Assessment of an Algorithm’s Ability to Change Copyright Litigation

Nicholas J. Isaacson is a J.D. Candidate, 2021 at NYU School of Law.

Will music copyrights lawsuits soon be a thing of the past? If Damien Riehl and Noah Rubin get their way, that future may soon come to pass. The pair have created an algorithm to compose every combination of musical notes possible with the stated goal being: “to copyright every single combo in order to give it to the public so musicians and artists can use melodies without worrying about copyright issues down the line.” Unfortunately, this project, while novel, has a number of clear feasibility problems, primarily stemming from issues of authorship, originality, and formal legal requirements. Even if those are overcome, given the case law surrounding music copyright cases, it is not clear how this would impact litigation (or if it would at all).

It is unclear whether Riehl and Rubin satisfy the copyright requirement of authorship. The lack of clarity here stems from the lack of definition of “author” or “authorship” in the Copyright Act. Case law does shed some light on the subject: the court in Aalmuhammed v. Lee, 202 F.3d 1227 (9th Cir. 2000), held that the “author” is to whom work as a whole owes origin, the “master mind” who exercises control. This definition would be crucial to Riehl and Rubin’s case as the algorithm would create the copyrightable material, not them, so they would have to argue that since they control the algorithm, they are the mastermind. They will need the court to accept this rationale; otherwise, the author is the algorithm. This outcome could be problematic. Naruto v. Slater, 888 F.3d 418 (9th Cir. 2018) established that an author must be a human although that case was about whether a monkey could be considered an author. The case law surrounding copyright of works created by algorithms and artificial intelligence is unsettled. If they can establish the algorithm is a tool, they are on safe ground. If not, the case gets complicated. Getting a court to embrace a friendly definition of authorship will be a struggle for Riehl and Rubin.

Originality is also required for copyright, and the compositions created by the algorithm arguably lack that. Originality, like authorship, has no statutory definition, so a definition must be sought from case law. Helpfully for Riehl and Rubin, the legislative history of the 1976 Copyright Act stated that originality is not novelty, ingenuity, or esthetic merit.[1] Unfortunately for the duo, independent creation requires a modicum of creativity. While that standard is quite low, the bar is not met because by creating every possible combination, the algorithm makes no choices, which are required for creativity. Burrow-Giles Lithographic Co. v. Sarony, 111 U.S. 53 (1884). The lack of creativity means that the works are effectively just non-copyrightable facts, which will stymie the Riehl and Rublin scheme.

While relatively minor compared to the last two issues, Riehl and Rubin will also have to deal with the formal requirements of copyright. For the copyright to have full force of law, the pair will have to register, deposit, and give notice of all of the works. On an individual level, this is not difficult, but this will become incredibly cumbersome when it must be done for every possible combination of musical notes (which will undoubtably be a very high number). All of this paperwork will take time and costs money. A standard copyright registration costs 35 dollars, which adds a hefty cost multiplier to registration efforts. This is not even taking into account the cost of giving two copies as a requirement for deposit. While the algorithm may create the works in a cost-effective way, the formal requirements required under copyright law will raise the costs substantially, if not exponentially.

Assuming all of the requirements of copyright are satisfied, it is still not clear what kind of impact the algorithm’s output would have on litigation. For copyright cases, there are three types of copying: de minimis, substantial similarity, and exact copying. Riehl and Rubin aren’t concerned with exact copying, but with similarity. Unfortunately for them, given the legal framework, having an algorithm compose every possible song won’t solve this. Their first problem is that compositions and recordings have separate copyrights, so the compositions are only half the problem. The second problem Riehl and Rubin will face is the innocent infringer problem, as was exhibited in Selle v. Gibb, 741 F.2d 896 (7th Cir. 1984). There, despite near-exact similarity, because there was no access, the court ruled there was no infringement. If the algorithm is publishing thousands and thousands of compositions somewhere online where no one ever listens to them, authors of new songs will effectively be able to infringe without consequence as they have no clear notice or access. And then they will be able use that copyright to sue in the manner Riehl and Rubin are trying to do away with. Finally, the algorithm’s new compositions will not eliminate copyright for the countless musical works that currently have valid copyright (with all the legal rights and protections that come with that). All of that being so, the impact of the algorithm looks like it would be minimal.

Riehl and Rubin’s algorithm plan is ambitious, but unfortunately that seems like all it is. They will struggle to be deemed authors of the algorithm’s work product. That same work product will likely not be considered to have the modicum of creativity necessary for copyright. Even if copyright is attained, the formalities of copyright will add a substantial cost multiplier to the endeavor. If all of these obstacles are overcome, it seems likely the algorithm’s output will have little to no outcome on copyright litigation. Riehl and Rubin’s plan seems unlikely to live up to its lofty ambitions.

[1] H.R. REP. NO. 1476, 94th Cong., 2d Sess. 47, at 51 (1976).