(This article is written by Shashank Tripathi, III-Year Student at The Rajiv Gandhi National University of Law, Patiala)
INTRODUCTION
Imagine losing an art competition to an artwork made by an Artificial Intelligence program that turns lines of text into hyper-realistic graphics. Human artists were left infuriated when the Colorado State Fair awarded Jason M. Allen with a blue ribbon for emerging digital artists for an AI-generated artwork. AI is permeating every domain of human action and is moving towards autonomy wherein it is capable of producing and performing tasks with close to minimal human inputs and assistance. As the evolution of every technology poses legal qualms, one of the many challenges AI poses pertains to Copyright Laws.
U.S. District Judge Beryl Howell observed in one of his judgments how, “we are approaching new frontiers in copyright as artists put AI in their toolbox,” which will raise “challenging questions” for copyright law. Effectively, AI increasingly replaces human involvement and causation in the creative process, the fundamental question of whether or not should Indian Copyright Laws grant protection to AI-generated works takes center stage.
This article will analyse the interaction of AI-generated artworks with the traditional copyright principle of human-centered originality in light of ethical questions posed by such works. Moreover, it will critically examine the current Indian legal framework – specifically Section-2(d)(vi) of The Copyright Act, 1957 – and its limitations in an attempt to provide for legal and policy changes to accommodate evolving needs.
HOW DOES AI GENERATE WORKS?
To examine the idea of authorship with respect to AI, we first need to delve into how AI produces final results. Ian Goodfellow, a prominent computer scientist and one of the creators of generative AI while explaining the potential of AI said, “You can think of generative models as giving artificial intelligence a form of imagination.” They typically learn to create new results by processing massive amounts of data, looking for patterns to model in their decision-making and, usually can do so without supervision through their analysis of various patterns of behaviors and commands. Scholars usually term this autonomous capability of generating outcomes as ‘algorithmic creativity’.
AI is not only capable of creating results based on pre-existing data but is evolving to make its predictions on data patterns thus allowing it to create works completely detached from human authorship. For instance, Google engineer Alexander Mordvintsev developed DeepDream, an AI tool powered by neural networks, capable of detecting, processing, and enhancing patterns of an image to create psychedelic results all on its own.[A2] Additionally, ChatGPT, an Open-AI creation capable of producing human-like text outputs in the form of poetry, prose, scripts, and codes, and DALL-E and VALL-E, the art generator and text-to-speech tools respectively, have pushed the limits of generative AI where it can assume the role of author to interpret inputs and generate outputs based on predictions rather than existing data. Thus, the outputs generated by these AI tools have an element of creativity that otherwise could be awarded a copyright in case of human authorship and therefore, they demand an analysis in light of the Indian Copyright Framework which has exhibited a traditional approach so far.
IMPLICATION OF GENERATIVE-AI UNDER INDIAN COPYRIGHT FRAMEWORK
Copyright laws intend to incentivize authors by protecting their work by limiting its reproduction without their permission. The authors are awarded with exclusive legal rights over their work to distribute and allow the use of their work in the public domain. However, the award of copyright under Indian copyright laws is subject to the originality of the work wherein the degree of originality decides whether a work shall be awarded copyright or shall remain open to the public. While Section 13 of the Copyright Act limits the award of copyright to, “original literary, dramatic, musical and artistic works”, it does not provide objective criteria to determine the originality of the work and it is subject to the courts to determine whether the work qualifies as original or not.
Consequently, the Indian courts have largely relied on the Doctrine of Sweat of the Brow expounded in the case of University of London Press v. University Tutorial Press to determine the originality of the work. In the given case, the court had established that as long as a work is not duplicated from another pre-existing work, and results from a modicum of ‘skill, labor or judgment’, copyright will subsist. Going by this doctrine, AI-generated works qualify for copyright ownership and mere reference to existing works would not equate to duplication. Since AI is inherently incapable of duplicating the work as algorithms are characterized by their capability to generate new works, a challenge may lie merely on the grounds of the absence of ‘skill, labor, and judgment’ as AI is capable of generating response within seconds which leaves doubtful grounds for the requirement of skill or labor.
Another layer of challenge to AI lies in its absolute reliance on the existence of prior works to process and produce new works. In other words, while humans do not necessarily need a pre-existing data pool to generate an original work, AI on the other hand is trained to generate responses after analysing and studying a large pool of available copyrighted works thus unauthorisedly using the copyrighted works to train itself. However, recently the US Court ruled in the case of Warhol Foundation v. Goldsmith, 2023 that while creative works deriving value from original works may not be treated as fair use, the limited usage of original work as long as it is meant for legitimate purposes while maintaining the sanctity and originality of the work is excused. Accordingly, interpreting Section 52 of the Copyright Act which excludes the reliance on original work for research purposes from the claim of copyright infringement, it may be construed that if the AI tools strictly rely on pre-existing databases to train their algorithms and do not produce results that derive value from existing work, may be exempted from the claim of infringement on the grounds of fair use.
Lastly, the Copyright Act defines “person” to include natural persons only and copyright cannot be awarded in the absence of evidence of individual involvement in terms of creativity. While generative AI is capable of generating responses without reliance on original works, the outcomes it produces are unforeseeable and thus, there is an absence of “individual involvement” to award it with an adequate originality threshold. Therefore, copyright cannot be awarded to an AI-generated work under the Indian Traditional Originality Framework. We may however seek recourse in the 1994 Amendment to the Copyright Act which included computer-generated works within the ambit of Copyright.
SECTION 2(d)(vi) OF THE COPYRIGHT ACT, 1957
Section 2(d)(vi) states that: “in relation to any literary, dramatic, musical or artistic work which is computer-generated, the person who causes the work to be created” shall be the “author”. The claim of copyright thus relies on the scope of “personhood” under the Copyright Act, 1957 wherein the authorship is limited to a natural person. When the amendment was brought in 1994, human involvement could be traced behind a computer-generated work. GenerativeAI on the other hand does not rely on human input and therefore, several authors may be traced for a given work. For instance, it could be argued that the software developer who developed the algorithm of the AI caused the work to be created and therefore, can be considered to be the author; it may also be the person who made the input to generate distinct results from the AI. In certain situations where the algorithm is developed such by the developer that a minor involvement is required by the one making the command to create results, joint ownership may be assumed for collectively causing the creation of the work.
However, determining the award of copyright on AI-generated work is not as straightforward as may be assumed owing to the large number of stakeholders involved in the creation of outputs. For instance, various image-generating AI tools are trained on millions of images, and data providers, developers, and machine operators among many others are involved in the development of such AI. Therefore, in regards to ownership, there is an ambiguity in Section 2(d)(vi) regarding the objective evaluation of what constitutes “causing” the work to be created. Additionally, even if joint ownership is treated as a plausible solution, proportional allocation of ownership shall cause a legal challenge that would cause more conflicts over title than it would settle.
The most problematic part of the provision lies in the presupposition that only a natural person may cause the work to be created from a computer resource while completely disregarding the autonomous nature of generativeAI. Consequently, while computer-generated works with substantial human involvement are protected under the Act, autonomous works of AI are beyond the scope of the provision. Whether or not the Copyright Act needs an amendment to introduce the provision for the inclusion of AI rests on the desirability of the copyright for AI-generated works.
WHO MUST OWN THE COPYRIGHT?
With the rapid rise of the digital economy, Governments would want to promote AI developers and incentivize them to solve more problems using AI. Without the grant of copyright, which awards the developers and users with economic rights, innovation and investments in the AI industry may dip. However, there are three limbs to the idea of an award for copyright: Whether the developer of AI, the creators, or the AI itself should own the exclusive rights over the results?
With regards to the developers, they are already awarded the rights over the software and database under the current Copyright framework. To award them with copyright over the work generated by AI would be an over-protectionist provision wherein they would own everything that the AI would be capable of generating. Moreover, since AI is capable of producing an infinite amount of work, to award copyright to the developers over every work would be disproportionateto the incentive that copyright seeks to provide.
On the other hand, reward theory scholars believe that the creator must own the rights over the work owing to the value they add to society through their work. This theory nonetheless has two drawbacks when pitted against generative AI; firstly, that generative AI has minimal to no involvement of the creator in the outcome and the right may not be justified on the grounds of investment of ‘skill and labour’, and secondly, companies and developers may restrict the access to AI owing to the profit it would generate for the countless copyright owning creators, thus, leading to anti-competitive software monopolies.
Finally, the Indian Copyright regime is based on human authorship and operates on the principle of incentivizing as well as protecting ‘human’ creativity. Contrary to this, an AI or Machine does not need an incentive to produce works, and neither do they possess the intellect, or capability to claim authorship or rewards of their work. Therefore, the conflict of copyright claim must not treat generative AI as one of the potential claimants for it is unnecessary in light of IP’s inherent purpose.
CONCLUSION
Coming back again to Colarado State Fair, when pitted against traditional artists, AI might be devaluing human creativity. An award of copyright on AI generated work, therefore, would lead to either the ownership majorly resting with programmers and developers, or with creators who lack skill and labour in their respective works. Either way, it would bid traditional artists against AI which inherently devalues creativity and labor behind human-authored works. Therefore, the most feasible solution to the issue, while not compromising on incentives to human authorship, would be the entry of AI-generated works into the public domain.
The primary result of the same would be a value addition to society without the monopolisation over AI generated works. For instance, if the developer claims rights over generated works, it would lead to limited usability leading to lower value addition and economic benefits. In contrast, adopting a free-for-all approach would, on one hand, ensure the survival of human artists in the face of highly capable AI, while on the other, it would promote value addition.
Under the Indian Copyright framework, the AI-generated works lack the human standard of originality and Section 2(d)(vi) remains outdated to accommodate for them. The reducing element of individual involvement and increased autonomy of AI puts up challenges of over-rewarding and causing unequal and inequitable access to AI in case of copyright award. To sustain the balance between human creativity and AI, and to promote value addition in terms of artistic ventures and economy, it is thus suggested that public domain be the solution to the conundrum and limitations of the current copyright framework.