Introduction:
The Colorado State Fair held its yearly art contest in the late 2020’s, which led to a strong disagreement between two opposing sides. Jason M. Allen won first place for his artwork Théâtre D’opéra Spatial, which he created using Midjourney, a generative artificial intelligence program to generate his artwork. The artists who protested against the victory declared it to be a form of dishonesty. The ethical debate about machine-generated images as “art” status actually hid a deeper problem that dealt with intellectual property law.

The basic problem involves determining who possesses the rights to ownership. Is it the person who creates an AI-generated masterpiece who will receive exclusive rights to the work or one who provides the AI with instructions will receive the rights or even the AI itself or the many developers whose work helped create the AI system? The current dispute shows how technology develops at a fast pace which conflicts with copyright rules that require authorship to come from human creators.

The legal gap emerges because we need to examine how Generative AI systems achieve their function. These systems utilize self-supervised training methods on extensive web-scraped datasets which include text, images, audio and video data. The system generates new outputs through its process of synthesizing data instead of producing exact copies of information.

The article examines legal complexities which arise from international copyright laws because both India and the United Kingdom provide statutory flexibility but the United States requires human authorship as its legal standard.

The US Copyright Office requires copyright protection to reward authentic human creative abilities while both the UK and India use legal fiction. The liberal frameworks grant authorship rights to any user who enters a prompt which results in devaluation of actual intellectual work achievements. Relying on these legal fictions allows a human user to acquire exclusive rights for work that a machine completed with little human involvement..

A fragmented jurisprudence:
The fast development of generative artificial intelligence technology has changed how people understand creative work, authorship rules and ownership rights. The increasing ability of AI systems to create text, images, music and audio-visual content which matches human production capabilities has created a copyright challenge which the legal system has never faced before. Legal systems built around the notion of human intellectual labour are now being forced to confront questions that were previously theoretical: whether training AI on copyrighted works constitutes infringement, whether machine-generated output can attract protection, and how far liability can extend across territorial borders. Courts in different legal systems are working to address these issues which has resulted in an incomplete yet developing body of legal precedents.

The United States: Doctrinal Clarity and the Strict Human Standard

Despite its pursuit of technological advancement, the United States clearly recognizes the significance of human-ness. Specifically, under Section 102(a) of the U.S. Copyright Act of 1976, any “original works of authorship” made by an author and fixed in a tangible form of expression is granted copyright protection. The Act dictates that such work must be perceivable, “either directly or with the aid of a machine or device,” which includes “all other forms of works,” including the literary works, musical compositions, dramatic performances, and pictorial arts. Despite what initially appears as an allowance for the use of AI in the realm of authorship by stating “with the aid of a machine or device,” judicial opinion holds consistently that the word “authorship” has always included the notion that creation is initiated by a person. A generative AI system allows human authors to work either with its help (e.g. Producing content using a tool that helps create) or for a speed increase in production time. The machine itself is not recognized as the author. Therefore, an author receives a copyright for only the human contributions to a work (those which demonstrate creative authority in selecting, arranging, editing or otherwise transforming what the machine has created) where a fully automated, machine-created work does not qualify for protection under Section 102(a).

The mandate that only people can create original works serves as the fundamental principle of human authorship. The Supreme Court recognized that a camera functions as a machine but a human being who makes all creative decisions determines the final product. The principle of human creative control which Burrow-Giles establishes remains applicable to present-day AI-assisted creative works.

The present-day American judicial system maintains strict enforcement of this legal principle. In the case of Thaler v. Perlmutter, the U.S. Court of Appeals for the D.C. Circuit determined that only human beings can create copyrightable work because ownership of an AI system does not grant rights to its generated content. The ruling used both statutory analysis of the US Copyright Act and constitutional understanding of copyright functions as a reward system for human creative work.

The unyielding doctrinal definition leads to two outcomes which create a fundamental policy conflict and an economic disparity. Purely AI-generated works immediately enter the public domain, regardless of their commercial or cultural value. The strict protection system establishes a protection gap which results in two problems: it minimizes important economic results and creates unpredictability for markets which depend on AI-generated content. This commercially unappealing result sustains copyright’s fundamental principle which grants exclusive rights to authors who create original work through their intellectual efforts rather than through computational capacity.

The United Kingdom: Statutory Flexibility and Legal Fiction

The United Kingdom uses a system of statutory attribution for determining authorial ownership of computer generated works, this puts it among a select few countries worldwide who have legislation on computer-generated IP. All intellectual property in the UK is governed by the Copyright Designs and Patents Act 1988 (CDPA), the Act creates entirely new legal categories for computer generated IP by classifying them as their own unique classification of IP. The author of computer generated literary, dramatic, musical or artistic work would be that of “the person by whom the arrangements necessary for the creation of the work are undertaken”. This proves that the computer or AI program cannot be the author, as it requires the presence of human activity to establish legal authorship. The law does not specify what “creative” decisions were made or by whom they were made but it specifies who used the computer program and what arrangement of tasks were made, the “arrangements necessary for the creation” being a widely interpretative phrase. The notion of who holds rights to the IP has changed from who wrote the code to what actual intellectual property created the final output. A computer generated work is granted protection for fifty years from the end of the year in which it was created compared with a human generated work which will be granted protection until fifty years after the author has died. The section still holds value in its current state for simple function computer programs but current AI systems work on stochastic processes to achieve output therefore the statute cannot comprehend current AI systems.

AI training methods have faced rigorous and often complex judicial interpretation by UK courts despite the statute’s ambiguity as to the concept of authorship. The UK High Court in Getty Images (US) Inc. V. Stability AI Ltd. Provides an interesting ruling for this debate. Getty claimed Stability AI’s AI model called ‘Stable Diffusion’ had infringed their copyright within the CDPA 1988 by being trained on millions of copyrighted images without prior consent. Getty produced a new theory that the AI system itself had caused secondary infringement and thus constituted an “infringing article”. However, this argument was rejected as although the digital file may technically be an ‘article’ it did not contain copies of Getty’s photographs but instead comprised of learned statistical weights that were then applied to creative inputs, this part of the legislation created a key distinction that can protect creative work as distinct from analytical models. In relation to Getty’s trademark case, however, the AI system was found guilty for producing images with Getty watermarks thereby misleading consumers according to the Trade Marks Act 1994.

The UK courts have consistently supported the idea-expression dichotomy which is a fundamental principle in copyright law. The Court in Nova Productions Ltd v. Mazooma Games Ltd declared that the source code and the art assets are to receive protection but the concepts and styles which make up the functionality will not be given such rights. This has very important implications for generative AI that imitate generic styles as they will be unlikely to receive such legal protection.

The Indian Imperative: Trapped Between Fiction and Doctrine

The current Indian copyright system maintains a difficult balance between its existing legal terms and the new technological advancements that it must handle. The Copyright Act of 1957 establishes the legal framework for copyright protection in India. The UK system of attribution for computer-generated works functions because the statute contains no direct references to artificial intelligence.

The Overbreadth of Section 2(d)(vi):
The Copyright Act of 1957 defines a computer-generated work author as “the person who causes the work to be created” according to Section 2(d)(vi). Introduced through a 1994 amendment, this provision predates contemporary generative AI and was intended to address tool-based computing environments

The legal fiction established by Section 2(d)(vi) functions similarly to the UK’s CDPA system. The system requires human participation to create authorship status that enables human designation as authors despite computers producing the entire work. A person qualifies as an author through the current text if they start the work process and control the tool while delivering prompts that lead to the end result.

The current definition of causation serves as an excessive legal standard for assessing contemporary data-driven autonomous AI systems. The process of determining authorship through causation creates a risk of incorrectly assigning ownership rights to individuals who lack any actual link to the content that was created.

This statutory overbreadth creates significant challenges for India. The Indian copyright system requires before they can receive copyright protection according to the rules of Indian law. India’s Supreme Court set a high bar in Eastern Book Company v. D.B. Modak. It declined copyright protection for mechanical work by stating it required originality based on skill and judgment along with a modicum of creativity. The system necessitates originality in all forms of work worthy of protection by it. Section 2(d)(vi) requires an actual work of creation be made by a person to fulfill the provision.

Eastern Book Company’s essential argument does not change because its doctrinal formulation centers on active human intellect as a component of authorship. Generative AI output is generated based on statistical inference not by actual human intellect. Using a fiction the provision of Section 2(d)(vi) merely gives actual causation as authorship, by foregoing any necessity for an actual work of creation. The section 2(d)(vi) system relies on merely controlling outputs through prompting, not in reality by controlling the outputs. Legal recourse becomes a tool to obtain property not actual intellectual work. As far as India is concerned no authoritative decision from any court addresses whether AI generated work deserves legal protection leading to a state of ambiguity and favouring big tech who are adept at leveraging legal ambiguities. The provision of section 2(d)(vi) is essentially a dormant rule which could have a broad ramification as there is so much scope to interpret the provision however deemed appropriate.

The Training Data Crisis and Jurisdiction

The ongoing authorship dispute in India prevents the establishment of proper AI training standards because it creates a fundamental structural flaw in the country. The operation of generative AI systems needs extensive computational resources to handle large datasets which frequently contain copyrighted material. The Copyright Act in India does not provide any specific legal exception for text-and-data-mining (TDM) purposes. Section 52 of the Act defines “fair dealing” exceptions through an exhaustive list of permitted uses that researchers can use for private study, criticism, and reporting purposes. The effort to classify AI training as “research” under current regulations creates legal uncertainty because private companies pursue AI training for commercial goals. The absence of a statutory TDM exemption creates a chilling effect on domestic AI development. The current situation requires companies to use foreign models which prevents India from achieving its goal to become a worldwide center of AI excellence.

The first of its kind legal case to actually test this intersection of copyright and generative AI is ANI Media Pvt. Ltd. V. OpenAI Inc. The High Court of Delhi examined whether ANI’s subscription-based news content was being used to train the language model without authorization, thereby infringing upon the exclusive rights of reproduction and communication to the public afforded under section 14 of the Copyright Act, 1957. ANI claimed that the false identification of its website with an AI-generated output and the misinformation it led to caused defamation.

The case is significant because it establishes a crucial test to determine the venue of litigation; OpenAI sought to have the case dismissed in India by arguing the model was trained on servers located outside the country, but the Delhi High Court relied upon section 62(2) of the Copyright Act which permits an author or owner of the copyright to initiate proceedings at their place of business. This robust interpretation reflects a strong stance by India to seek jurisdiction in an international case concerning the legal rights of AI systems. The High Court rejected an interim injunction primarily because OpenAI had already blocked ANI’s website, but the implications of the lawsuit are far-reaching. This instance demonstrates that while section 52 already seems flexible enough to allow AI training as fair dealing, there is a clear need for India to have dedicated regulations.

There is a need to develop a legal solution within Indian copyright law that resolves the issue between the legal fictions introduced by amendments of 1994, and the contemporary technological development. India’s copyright law has always had the intention of protecting the authorship and dignity of an individual human being in relation to their artistic work and its cultural labor. Attribution of authorship to those who “caused” the AI output undermines traditional relations between author and work, and in so doing undermines the very framework of copyright law.

Conclusion:
The examples of the United States and the United Kingdom clearly demonstrate the need to formulate an Indian system of tiered protection under certain limitations. Legislative amendment is necessary for section 2(d)(vi) so that its application doesn’t extend beyond what is reasonably imaginable in copyright law. The text-and-data-mining exception under section 52 of the Copyright Act needs to be fleshed out with an appropriate regime to prevent market substitution while simultaneously allowing research. An exception that limits the applicability of moral rights within an exclusive economic regime will further allow for corporate control over highly valuable, autonomously generated AI output. Indian copyright law needs a framework such as this for its continued efficacy.

The underlying problem in establishing legal rights with regards to generated AI output is identifying the true creator of the art product. An AI system works by processing datasets that contains copyrighted material; it lacks the ability to interpret or discern aesthetics. Statistical inference, on which the entire process relies, is fundamentally opposed to the foundational principles on which copyright is built. The issue is the conflicting demands of consistency and market needs of global legal systems.