Blurred lines of A.I. require musical clarity (2/3)
Since then, many other instances of AI-generated songs have spread across the internet, replicating the voices of singers including Harry Styles, Rihanna, and Kanye West.
Hence, the first big question appears to be: does AI violate artist copyrights? The second big issue is, what rights do human copyright owners have when AI creates something? And here, there are two categories of questions — an input question and an output question. On the input side, does the training that is required to create these complex AI models infringe copyright? In other words, if AI is trained to extensively listen to music, has the copyright of the owners of that (piece of) music been infringed if done so without their consent? Or is that protected in some manner by way of “fair dealing” under Section 52 of the Indian Copyright Act, 1957?
Then there’s the output question, which is, if copyright law provides the owner the exclusive right to create a derivative work based on prior work, is creating something using AI based technology on that “other” work result in copyright infringement, or can that “other” work be considered as “original” work and, hence, registerable in its own right?
In general, music in the style of someone else is generally not considered as derivative work for the purposes of copyright law and may be permitted but, by utilizing machine learning and AI-generative work, it remains moot whether those outputs themselves are protected. So, both the input and output questions remain both unresolved and complicated.
On March 15, 2023 the U.S. Copyright Office issued a formal guidance, which reaffirmed its position that works that are created by AI without human intervention or human involvement cannot be copyrighted.
For the record, for any work to be accorded copyright protection, it must meet a few essential criteria: the work must be a work of human authorship, and the work must be “original”, as can be gauged from Section 13(1)(a) of the Act. Further, the Copyright Office may always refuse to register works that have been generated with the use of an AI tool on the grounds that it lacks sufficient original creative input, and human authorship necessary to support copyright legitimacy. Hence, the issue of copyright protection of AI-generated content has two facets: First, whether AI can be named as the author; and, second, whether such AI-generated content has sufficient creative inputs to meet the criteria of originality.
In leveraging the position of Indian laws governed by the Copyright Act, 1957 one of reasons as to why, in India, a copyright may not be granted to AI is Section 2(d), which defines that, for validation of ownership of any copyrighted work, the person should fall under the domain of an “author”, which has been interpreted repeatedly to mean a human as the Act also makes reference to the entity being “a citizen of India”. So, as a corollary, the relevant sections make it abundantly clear that only human beings can be authors for purposes of constitutional and statutory copyright purposes.
This obviously bars AI from owning any authorship because AI cannot be regarded as a legal person but, at the same time, Section 2(d)(vi) discusses authorship in relation to any literary, dramatic, musical or artistic work “which is computer generated” as “the person who causes the work to be created”. If this part of the Act, read with the definition of “computer programme” which means “a set of instructions expressed in words, codes, schemes or in any other form, including a machine readable medium”, may still give rise to the argument that AI-generated content is copyrightable.
However, these arguments still require to be fully tested in courts and it appears to be only a matter of time when they will be as, based on the previous arguments and the interpretation of the Copyright Act, 1957 AI-generated material may still be copyrightable wherein there is sufficient human authorship by any person who has selected or arranged AI material in a creative manner.
All this leads to yet another question: can a voice be protected by copyright or any form of intellectual property (IP)?