Intelligent machines are here, but they do not come as conquerors, they come as creators. From designing websites and writing essays to producing code and generating artwork, there is little Open AI’s Chat GPT, a state-of-the-art artificial intelligence chatbot, cannot do.
And while human efforts to train computers to produce art have been around since the 1970s, this time it’s different. Early machine generated works heavily relied on the creative input of the programmer. The machine was simply a tool for the artist, much like a musical instrument or a paint brush. This is no longer true. Machine learning software, a subset of AI, can generate autonomous systems capable of learning without specific human programming.
Programmers may still define certain parameters, but the artistic work is created entirely by the programme through its neural network: a process much like the thought processes of humans. When applied to music and literary works, AI machines make entirely independent decisions in its output. Doubtless, this creates wider implications of ownership and infringement under copyright law.
Take ownership. Who owns copyrights in art generated by AI? There are principally two approaches possible where human control is minimal or entirely non-existent. The law may either deny copyright protection entirely or attribute authorship to the creator of the machine.
The first approach is followed by much of the West. In the USA, for instance, the Copyright Office declares that it will “register an original work of authorship, provided that the work was created by a human being”. The European Court of Justice in its landmark decision (C-5/08 Infopaq International A/S v. Danske Dagbaldes Forening), held that copyright only applies to original works, and originality must reflect the “author’s own intellectual creation.” Jurists have understood this to mean that a human author is a prerequisite for a copyright work to exist.
Conversely, India, among others, has followed the second approach. Under Section 2(d)(vi) of the Indian Copyright Act, an ‘author’ means “in relation to any literary, dramatic, musical or artistic work which is computer generated, the person who causes the work to be created”. Still, Indian authorities are unclear on how to implement this definition. In 2020, the Indian Copyright Office rejected an application which listed AI (RAGHAV) as the sole author of an artwork. The Copyright Office subsequently granted an application where a natural person along with AI (RAGHAV) were named as co-authors, and then shortly thereafter, issued a withdrawal notice. The applicant has naturally challenged the withdrawal notice and the appeal so far remains undecided.
That apart, the Indian definition fails to address a fundamental issue: does the law recognize the creator of the programme or its user as the individual who ‘causes the work to be created’? Put differently, should copyrights be conferred on the maker of the pen or the writer? Microsoft, for instance, does not own copyrights to work created on Microsoft Word. In the context of AI-generated work, text-generating machine learning programs can produce a sonnet that resembles the language of Shakespeare with the input of a simple set of instructions. The user’s contribution may not be much more than pushing a few buttons.
How do courts address such complexities? So far, the solution has been fashioned on a case-to-case basis. In the English case of Nova Productions v. Mazooma Games Ltd and Ors.  EWHC 24, for instance, the English courts were asked to determine authorship of a particular display in a computer game. The Court noted that the appearance of any screen or display in a video game depends to some extent on the way the game is played. But a player’s input “is not artistic in nature and he has contributed no skill or labour of an artistic kind” and no copyright ownership can thereby extend to the player.
The copyrighting is on the wall
More importantly, AI-generated artwork create substantial concerns of infringement. Today, entire oceans of copyright protected data is consumed by bots to create human-like expressions of art. This data is plundered from everywhere: digital libraries, internet databases, and social-media feeds. Those responsible for protecting source material allege that data is siphoned up without any consent or credit. In effect, the argument goes that machines are infringing copyright law in the act of learning. Critics say creative industries should ensure artists’ works are licensed and the technology industry should act ethically in training AI.
Programmers conversely argue that training AI should be protected by the ’fair dealing’ doctrine and data mining without licences should be permissible. Under Section 52 of the Indian Copyright Act, a fair dealing with any work for the purpose of (i) research or private study, (ii) criticism or review, or (iii) reporting current events, shall not constitute copyright infringement. Under 52(i)(i), this includes “the reproduction of any work by a teacher… in the course of instruction”.
Conceptually therefore, usage of copyrighted material for training purposes may be protected by law. More so, in cases where use of data serves a valuable social purpose, and the material is transformed from the original without affecting the owners’ core target market.
Authors of AI also insist that access to a wide array of training sets is fundamental to improving AI, without which, the industry may die before it truly starts. They frame it as a fundamental question that faces humanity: will copyright law allow robots to learn? It is still early days, but the first lawsuits have started to fly. Getty images, a photograph agency, has accused Stability AI of infringing its copyright in millions of photos to build an image generating model that will compete with Getty. Assuming the case is not settled out of court, it may provide the jurisprudence necessary to address these legal complexities.
Still, these issues raise broader questions on the nature of knowledge. Not all things are covered by copyright. Most, if not every work is, to some extent, influenced or derived from other works. To prevent the consumption of copyrighted material for creation of art that may genuinely be different will transform the law into an instrument of oppression rather than an incentive for creation. Copyright protection will then move not only to covering artistic works inspired by others, but to ideas themselves.