Posted on November 4, 2020
Authored by Anunay Kumar Sharma*
In 2019, Warner Music, the 3rd largest music label in the world with artists such as Led Zeppelin, Ed Sheeran, Madonna signed on – not an artist, but a bundle of code – Endel. Endel is the byproduct of a startup based out of Berlin which consists of an Artificial Intelligence (“AI”) software which creates music based on the time of the day, weather, location and the music choices which it has studied.. The contract has been signed for up to 20 albums to be released on a 50-50 distribution deal. This is not the first time that software has been used by record labels, Sony has been working on its ‘Flow Machines’ project which uses an algorithm to compose music and has been credited officially as one of the songwriters along with several people who feed in the lyrics and different inputs, Aiva Music was the first AI to be registered with a collection society (SACEM). The point of difference between these previous occurrences is that Endel does not require any human collaboration apart from creating the AI itself. No human editorial control is required during the creation phase therefore the AI is not just generating music but creating the sound frequencies at the same time. The system was created by Dimitri Evgrafov and he explains that what people might think as music being created at the click of a button is the culmination of 1.5 years of developing the algorithm.
Can AI Be Credited With Authorship? The Philosophical and Legal Stance
This is an issue which arises organically out of any creative work being created, who will be attributed as the author? Since the early 1970s, when the use of software to create pieces of art was rising, it made sense to attribute the person making the art or controlling the machine to be attributed as the author and owner of the art while the machine or the computer was more of a tool to be used for creation. A similar decision was made by Warner Music and Endel, the owner of the company and the engineers who built the system have been credited as song writers even though hardly anyone of them know how to make music. It does make sense to attribute ownership of the music coming out of a system to the creator of the system. But, it gets interesting when the system is self-learning and after creating a few pieces of music with the help of the developers, it starts making something which the creators themselves hadn’t even thought of. In that case, a pertinent question arises – would the owner of the company be credited as songwriter for perpetuity for every piece of music coming out of Endel? If coders and engineers are credited as composers and the use of AI will only increase, at what point of time will the majority of song writers consist of coders rather than actual musicians? Attributing authorship to AI also leads to the problem of who will be liable in case of infringement. For instance, if a software studies Beyonce’s music and creates something which infringes the original work and the authorship has been attributed to the software itself, then who would the artists like Beyonce or the record labels holding the title to her music look to sue? The engineers can argue that as the creation of the song did not require any inputs from their end and thus question how they can be held liable. Thus, AI is slowly moving away from being just a tool to create art but something capable of making independent creative decisions.
For AI to claim authorship it needs to be determined whether it can create original and creative work or not. Under Section 13 of the Copyright Act, 1957 (“Act”), the work needs to be original for it to be protected as a copyright work. Courts have come up with certain tests to decide whether something can be protectable as there is no definition of the word ‘original’ in the Act.
- If the work was created using skill and labor (Sweat of the Brow doctrine) [Feist Publications, Inc. v. Rural telephone Service Co., 499 U.S. 340 (1991)]
- If the work involves some form of creativity (modicum of creativity doctrine) [Eastern Book Company v. D.B Modak, (2008) 1 SCC 1]
If the AI is to be able to claim authorship over any art, it would have to pass through the above mentioned tests. Although so far, any AI software mostly works on the information already available to it, it can be argued that it merely compiles different sounds and information (although philosophically it can be argued that any music or art has been inspired by some preexisting art). Even if AI is merely compiling rather than creating it, it was decided in Eastern Book Company & Ors vs D.B. Modak & Anr. (2008) 1 SCC 1) that compilation works can also be copyrighted if it is shown that it involved some skill and judgment. Most of the music coming out today involves, to a large extent, compilation of different samples and sounds to create something new. If any software is doing the same, it does require some creativity and judgement to make something which sounds new from pre-existing pre-sets.
In countries like Australia (Acohs Pty Ltd v Ycorp Pty Ltd  FCAFC 16), and in Europe (C-5/08 Infopaq International A/S v Danske Dagbaldes Forening) it has been stated multiple times that only a human can own a copyright. Similarly, it was recently decided by the US Courts in Naruto v. Slator that a monkey or any other animal cannot sue for infringement of copyright and therefore cannot own a copyright even though they have a constitutional standing. Although no such statement has been made against a software, but it would be hard to argue in case of software when it has been refused for a monkey. For an AI to claim authorship or to own a copyright, it would have to be recognized as a separate legal entity or a legal person. Further, complications arise in terms of infringement by an AI. Under Section 51 of the Act, only a ‘person’ can infringe copyright and as an AI is not yet recognized as a separate legal entity it would create problems for fixing liability on a particular person, considering the particular software did not require any human intervention for creating that material.
Under the Act, there are certain rights which are attributed to an author which would lead to further complications in case of an AI claiming authorship. For instance, Section 17 states that an author is the first owner of the copyright and therefore can transfer the ownership of his or her work, this would become complicated as an AI cannot authorize or execute any transfer (yet). Furthermore, Section 57 pertaining to moral rights including rights to paternity and integrity, these rights may become redundant with an AI. An AI system may not be able to ascertain whether its honour or reputation has been affected or not. Another issue which might arise is with licensing its work, who will be responsible for determining the licensing deals and royalty structures?
The Way Forward
An AI generated work can be divided into two broad categories – works created by AI involving human input and works created without human input. In case of the former, the ownership of the work can be attributed to the person who manages the input and the authorship can be shared by the AI as well as the person. In case of the latter where AI works independently, then the authorship can be solely attributed to the AI, but the ownership can be attributed to the person who created the system or the company which owns the system. In such a way, in case of infringement, there will be a clear entity for people to go after. For this to happen, the Act will have to recognize AI as a separate entity or the work being created by an AI will have to be classified differently. It can be argued that the work created independently by AI can be given authorship as any AI (working at the moment) operates based on certain parameters, codes and previous information provided by the developer, as the system evolves it creates new work based on those parameters and information even if the work involved compilation instead of pure creation it involves a certain level of skill, judgment and creativity to come up with something new.
No legal system in the world at the moment is equipped well enough to tackle the issue of AI creating work independently without human input. In countries like Australia, Spain and Germany, it has been briefly discussed to classify all the art created independently by any AI as open-source as no human factor is involved in any of those pieces of art. But, this leads to another problem, as why would investors and corporations develop some program if anything created by them belongs to the public domain? The world of AI is changing and evolving every day and new technologies, which felt like science fiction a few years back, are now a reality. Slowly, in almost every industry, the culture is shifting towards works created by AI in some form or the other and the legal discourse needs to appropriately move forward to create a balance of rights and liabilities related to the work created by AI.
Also see our ongoing ‘AI and Authorship’ series:
- “The Messiah or the Mirage: Analysing Shenzen Tencent Vs Shanghai Yinxun’s Contribution to the Discourse on AI and Authorship” by Aditya Krishna.
- “Yet Another Paradox of Schrödinger’s Cat: India and the Discourse on AI and Authorship” by Aditya Krishna.
*Anunay Kumar is a 4th-year law student at O.P. Jindal University and it was during the lockdown that he got interested in the field of IPR and decided to pursue it as his specialization