-Subhajit Basu* and Ankeeta Dutt**
Abstract
Artificial Intelligence (AI) systems have experienced rapid growth, permeating diverse sectors such as logistics, transportation, and healthcare, catalysing significant innovations. One particularly transformative area for AI has been the arts. While AI's integration into these fields demonstrates its vast potential to benefit society, its development raises critical legal and ethical concerns, particularly regarding using copyrighted materials without consent. Many AI companies utilise third-party creative works—such as literature, films, and art—as training data for their models, often bypassing the copyright holders' permissions. This practice not only undermines the rights of creators but also threatens the integrity of the industries that rely on these creative works. This paper critically examines India's copyright law, focusing on the provisions of the Copyright Act 1957 that pertain to the use of third-party works as AI training data. It assesses the adequacy of these provisions and underscores the urgent need for enhanced legal protections to safeguard creators and uphold the ethical boundaries of AI development.
Introduction
Artificial intelligence (AI) has evolved rapidly from a specialised technology into a fundamental component of various creative processes. Initially confined to data analysis and automation tasks, AI has emerged as a powerful tool for producing art, literature, music, and other creative outputs. This significant technological shift necessitates a critical re-evaluation of the socio-economic frameworks underpinning cultural production, particularly concerning the role of copyright law in promoting creativity and managing the complex effects of technological change on artists, authors, and the markets for their work.
The advent of AI-generated art has precipitated intense debate, particularly regarding its ethical and legal implications within intellectual property rights. As India becomes increasingly integrated into the global AI ecosystem, it is imperative to examine how Indian copyright law addresses the challenges posed by utilising creative works as training data for AI models. This analysis raises critical questions concerning the adequacy of current legal frameworks in protecting the rights of copyright holders when AI systems employ their works without authorisation. Moreover, it prompts a deeper inquiry into whether existing legislation sufficiently safeguards creators' rights or if further legal reforms are necessary to address the unique challenges presented by AI.
AI models such as Chat-GPT, trained on extensive datasets harvested from the internet, can generate text that closely mirrors human writing, often rendering it indistinguishable from content produced by human authors. Similarly, AI systems like Midjourney and Stable Diffusion can transform simple textual prompts into complex visual artworks with speed and efficiency far exceeding human capabilities. Unlike traditional digital art, which requires significant human creative input, AI-generated art typically requires minimal user involvement—often limited to a text prompt—from individuals who may lack artistic expertise.
The accessibility and ubiquity of AI-generated art have expanded significantly in recent years. Although AI-generated art is not entirely novel, its origins can be traced back to 1973 when computer scientist Harold Cohen developed AARON, a programme capable of autonomously producing abstract artwork. Within AI, the foundational building blocks are computing models composed of algorithms, neural networks, and statistical methods. Artificial neural networks, inspired by biological neural networks, are employed in 'deep learning' to enable computers to learn by example. When applied to art, these neural networks can generate artistic output. Among the commonly utilised neural network models are Generative Adversarial Networks (GANs) and Long Short-Term Memory (LSTM) networks.
The development of GANs in 2014 marked a significant advancement in AI's creative capabilities. Although GANs were not exclusively designed for artistic purposes, they enabled the creation of highly realistic images through the interaction of two neural networks—one responsible for generating images and the other for evaluating their authenticity. A notable example of early AI art was Google's DeepDream program, introduced in 2015. A key challenge with AI-generated art is that, while these programs can be trained to simulate visual perception by exposure to vast datasets, they lack the ability to comprehend how elements in the images they produce fit together meaningfully. This limitation was highlighted in 2018 when the commercial potential of AI art gained attention after Christie's auctioned Portrait of Edmond Belamy (Comte de Bellamy), an image created by three French students using a GAN. However, the release of OpenAI's DALL-E in 2021 further accelerated the expansion of AI-generated art. Trained on millions of images paired with textual descriptions, DALL-E facilitates the creation of unique artworks from simple text prompts, effectively bridging the gap between language and visual creativity. This breakthrough has encouraged the proliferation of AI art tools developed by private enterprises and open-source communities worldwide.
As AI continues to reshape the landscape of creative industries, the challenges confronting Indian copyright holders—whose works are increasingly being utilised as training data for AI models without consent—demand urgent attention. While existing legal frameworks offer some degree of protection, they may not fully address the complexities introduced by AI technologies. Consequently, it is essential to critically assess the adequacy of current Indian copyright law and consider whether additional legal measures are required to ensure that creators' rights are not compromised in this emerging era of AI-generated content.
AI-Generated Art in India
India has not remained insulated from global trends in AI art. The increasing prominence of AI-generated art on Indian social media, where "fantasy versions" of Indian landmarks and cultural icons are frequently depicted, highlights the nation's swift engagement with this technological phenomenon. In October 2021, Indian artist Raghava KK, in collaboration with AI artist Harshit Agrawal, neuroscientist Abhijeet Satani, and material scientist Ben Tritt, created La Petite Mort. This work, which integrated digital and AI tools, was an NFT accompanied by a physical artwork representing data and a visual interpretation of the human brain during an orgasm. In 2022, Raghava extended this exploration with the launch of Cyborg Desires, a project similarly driven by AI technologies.
The integration of AI into Indian art is further evidenced by artists such as Tapan Aslot and initiatives like the Times of India's "History meets AI" Instagram posts, which feature AI-generated images of significant historical events paired with informative captions. These examples underscore India's rapid adoption of AI art within the digital domain, paralleling international developments. Indian AI artists increasingly employ this technology to interrogate complex themes such as ecology, nature, gender, politics, and the human-machine interface. A notable instance is the AI-bot artist Auria Kathi, developed by Bengaluru-based designer Fabin Rasheed and Kochi-based engineer Sleeba Paul, which emerged in early 2019 with AI-generated profile images, poems, and abstract artworks disseminated through social media platforms. Harshit Agrawal's project Strange Genders, supported by the art research and curatorial collective 64/1, further exemplifies the critical potential of AI in art. Using GANs, Agrawal explored gender identity by training a GAN on a dataset comprising over 2,300 figures of male and female forms drawn by individuals from across India. The Indian government has also recognised the importance of AI education and accessibility. The launch of the "AI For All" website aims to democratise AI knowledge, while the Central Government's IndiaAI initiative seeks to enhance public access to AI-based functions and computation. It is underpinned by a dedicated research team committed to the widespread dissemination of AI-related information.
The popularity of AI-generated art in India presents significant challenges and opportunities, particularly concerning the nation's intellectual property framework. Although the Copyright Act of 1957 addresses computer-generated works, the issues of authorship and originality in the context of AI remain contentious (See Chakraborty and Gaffar). As AI models become increasingly sophisticated, determining the extent of human involvement in the creative process becomes more complex. This ambiguity may lead to disputes over ownership rights, licensing, and fair use. Additionally, the potential for AI to produce content that infringes existing copyrights raises critical concerns regarding intellectual property theft. To effectively navigate these challenges, India may need to reconsider its legal framework and provide clearer guidelines that balance the rights of creators, users, and AI developers in the evolving landscape of AI-generated art.
Copyright Law in India
The artistic community, particularly those involved in anti-AI activism, contends that AI-generated imagery constitutes a form of intellectual theft. They assert that AI systems unlawfully appropriate the creative labour of artists by using their works as training data without obtaining proper consent, thereby diminishing the value and rights of human creators. This concern is exacerbated because of the Indian Copyright Act 1957, which traditionally confers exclusive rights to creators for using and distributing their works, does not explicitly address the complexities surrounding AI-generated content. Historically, copyright protections have been anchored in human-generated works, including literature, music, and art. However, the emergence of AI technology, capable of producing original works with minimal human input, introduces significant legal ambiguity regarding the ownership of copyright in AI-generated art. The current lack of explicit legal provisions within the Indian Copyright Act leaves substantial gaps in determining the rightful ownership of works produced by AI, thereby challenging established concepts of authorship and originality. As AI continues to evolve, creating content that increasingly rivals human creativity, the issue of copyright ownership becomes progressively contentious. This legal uncertainty, coupled with the potential for AI to infringe upon the rights of human creators, underscores the pressing need for a comprehensive re-evaluation of existing copyright legislation.
Human vs AI: Complexities of Copyright Ownership
When discussing AI-generated works, two primary scenarios arise: works created by AI with human guidance and those created by AI without human guidance. In the first scenario, where humans provide significant creative input, guiding the AI to produce a specific output, copyright can generally be attributed to the human creator. This is similar to how a photographer retains rights to photographs taken with a camera. The human involvement ensures that the work meets the criterion of originality, as it is a product of the author's skill, judgment, and creativity. However, the situation becomes more complex when AI generates work independently, without direct human input. Attributing authorship to AI in such cases is challenging because current legal frameworks do not recognise AI as an entity capable of holding rights. For a work to qualify for copyright protection, it must be original—stemming from the author's skill and creativity. The originality of AI-generated works is contentious, as AI relies on pre-existing data and algorithms programmed by humans. For example, ChatGPT is trained on vast amounts of data, including copyrighted material, to generate text, and Google has developed software that can produce original music from descriptions and recordings. These technologies have the capacity to replicate and mimic existing copyrighted works, blurring the lines between what is original and what is AI-generated, thereby creating significant legal complexities. Determining whether AI possesses the necessary creativity to meet the threshold of originality remains a challenge, raising concerns about potential copyright infringement.
The Indian Copyright Act 1957 recognises only a "natural person" as an author. Section 17 of the Act specifies that the author is considered the initial owner of the copyright. However, as a non-human entity, AI does not fit within this definition and lacks the legal capacity to transfer rights or authorise ownership claims. This presents a legal conundrum, challenging traditional concepts of ownership and authorship in the context of AI as a creator. In an attempt to address these complexities, the Copyright Act was amended in 1994 to include computer-generated works, defining authorship under Section 2(d)(v) as "the person who causes the work to be created." Interpreting the term "person" is crucial in determining whether AI can be considered an author. Section 57 of the Act, which addresses the moral rights of authors—including the right to paternity and integrity—was designed to protect the personal and reputational interests of human creators. Applying these rights to AI, presumed to be an author, is difficult, as AI does not possess human emotions or subjective experiences. This raises significant questions about the relevance and practical application of moral rights in scenarios involving AI.
There is no legal precedent in India for recognising AI as an author, and the Copyright Office has yet to issue definitive guidelines. The Office's inconsistent rulings concerning copyright registrations for AI-generated works have exacerbated the prevailing uncertainty. In the recent case of RAGHAV, where both a human and AI were listed as co-authors in an AI-assisted artistic effort, the Copyright Office initially granted registration but subsequently retracted it. The Office contended that while works containing AI-generated content may qualify for copyright protection, such determinations would require a "case-by-case" analysis, focusing particularly on the functionality of the AI tool and its contribution to the final work. However, several inconclusive key issues remain, including the necessary level of human input required to qualify the user of an AI system as the "author" of a generated work, the extent of protection that can be afforded to the resulting image, and how to assess the originality of AI-generated works, particularly when these systems may have been trained on pre-existing works of unknown origin. Additionally, questions persist regarding how copyright law in India might best be utilised to incentivise creative works involving AI and other related complexities.
In a series of decisions (See “Zaraya of the Dawn” registration case and “Théâtre D’opéra Spatial” registration case) and guidance issued in 2023, the US Copyright Office has clarified that works created with substantial AI input are not eligible for copyright protection in the United States. The US Copyright Office Review Board reinforced this position by rejecting a request to register a two-dimensional artwork titled "SURYAST," generated by inputting a photograph into the RAGHAV app. Its position reflects a broader reluctance to diverge from the established principle that "human authorship is a bedrock requirement of copyright," as underscored by the US federal district court's ruling in Thaler v Perlmutter last year, which denied registration for "A Recent Entrance to Paradise," an entirely AI-generated image. The Review Board also drew on the Ninth Circuit's 2018 decision in Naruto v Slater, where it was held that photographs taken by a monkey are not protected under the Copyright Act, as the requirement for human authorship remains paramount.
Incorrectly granting copyright registration for AI-generated art can have significant implications. If such registrations are upheld, they could set a precedent for future cases, complicating the legal landscape and potentially misleading the public about the rights of artists. The question of who is entitled to royalties for AI-generated content further complicates matters. While current copyright laws under Section 18 of the Indian Copyright Act protect human authors' rights to royalties, the application of these laws to AI-generated works is unclear. Determining fair royalty amounts and ensuring equitable distribution is particularly problematic when AI is involved, as traditional mechanisms for assessing authorship and value may not apply.
The potential for AI to autonomously determine royalty amounts raises concerns about the fairness and reasonableness of such decisions. Moreover, the issue of AI accountability in content creation presents additional complexities. If AI generates harmful or offensive content, traditional legal remedies—such as removing the content or disabling the AI—may not sufficiently address the underlying issues of responsibility and accountability. Without clear guidelines and regulations, granting authorship rights to AI could lead to broader legal and ethical challenges. Developing a comprehensive framework that balances the rights of AI developers, content users, and the public interest is essential.[1] This framework should address issues such as ownership, licensing, and liability in the context of AI-generated content, ensuring that the evolving landscape of digital creation is navigated with clarity and fairness[2].
Case Law and Precedents in India
Indian courts have consistently favoured the recognition of human authorship in matters of copyright, reflecting a deep-rooted understanding that copyright law is inherently linked to human creativity, as shown in Rupendra Kashyap v Jiwan Publishing House Pvt. Ltd, the High Court of Delhi affirmed this principle by ruling that an artificial entity, such as the Central Board of Secondary Education (CBSE), could not claim copyright without demonstrable evidence of individual human involvement in the creation of the work—in this case, question papers. This decision reinforced the notion that copyright ownership is inextricably tied to the involvement of a natural person, thereby excluding non-human entities from the scope of authorship.
Similarly, in Navigators Logistics Ltd. v Kashif Qureshi & Ors., the Court's dismissal of a copyright claim over a computer-generated list further emphasised the indispensability of human intervention in establishing copyright. The Court's rationale was embedded in the traditional understanding that copyright protection is reserved for works that embody human skill and creativity. The absence of such human input in the generation of the list precluded the possibility of copyright protection, thereby highlighting the limitations of the current legal framework in addressing the complexities introduced by AI-generated content.
In the case of the Institute of Chartered Accountants of India v Shaunak H Sayta & Ors. Supreme Court offered further clarity on this issue. The Court held that question papers, evaluation instructions, and model answers—classified as literary works—are protected by copyright and are initially owned by their human creators. This ruling underscores the Court's emphasis on human authorship as a cornerstone of copyright protection, reaffirming the importance of human creativity in creating copyrighted works. The Court's recognition of the authorship of individuals, even within institutional frameworks, reinforces the legal doctrine that attributes copyright to natural persons.
However, despite these judicial decisions, the Indian legal framework remains ill-equipped to address the emerging challenges posed by AI-generated content. The Indian Copyright Act 1957 does not provide explicit guidance on the attribution of copyright for works generated without direct human input, nor does it clarify the legal status of AI as a potential "author." This lacuna becomes increasingly problematic as AI technologies advance, enabling content creation that may mimic or surpass human creativity. Applying moral rights under Section 57 of the Copyright Act to AI-generated works further complicates the legal landscape. Moral rights, which protect the personal and reputational interests of human authors, presuppose the existence of human emotions and subjective experiences.
The attribution of such rights to AI, which lacks the capacity for emotional or subjective experience, raises fundamental questions about the relevance and practical applicability of moral rights in the context of AI-generated content. The current legal framework, established to protect human dignity and reputation, is ill-suited to accommodate non-human entities that do not possess these attributes. Furthermore, the issue of accountability and liability for AI-generated content presents additional legal and ethical challenges. The traditional legal framework assumes that a natural person can be held accountable for creating and disseminating content. However, as a non-human entity, AI cannot be subjected to legal penalties or held responsible for content that may be defamatory, obscene, or otherwise harmful. This raises significant concerns about the potential misuse of AI in content creation and the adequacy of existing legal remedies to address such issues.
Ethical Debate: Is AI Stealing Human Creativity?
While AI indicates new avenues for creative expression, it concurrently raises profound ethical questions regarding copyright infringement and the preservation of human creativity. Critics contend that AI models are often trained on vast datasets from the internet, including images, music, and text—potentially incorporating copyrighted material without the requisite permission. This practice has engendered allegations of "digital plagiarism," wherein AI-generated outputs bear a striking resemblance to, or even outright replicate, existing works.
A primary concern in AI-generated art discourse is the extensive use of image datasets in training these models. Many of these datasets contain copyrighted works utilised without the consent of the original artists, leading to accusations of copyright infringement. Companies such as OpenAI and Meta have faced significant criticism for relying on such datasets, which is perceived as exploiting artists' labour without appropriate compensation. Efforts to mitigate these concerns have been somewhat limited. Certain AI companies have instituted opt-out programmes, allowing artists to request the removal of their works from training datasets.
Additionally, projects like "Glaze" have been developed, enabling artists to alter the pixels of their artworks to confound AI databases. However, these processes are often cumbersome and time-intensive. The sheer scale of these datasets further complicates the ability of artists to track and identify their works. Notwithstanding these challenges, websites like "Have I Been Trained?" have emerged, allowing artists to ascertain whether their works are included in datasets such as LAION-5B, a vast image library used in AI training. These developments underscore the ethical dilemma of whether AI art tools are unfairly capitalising on the intellectual property of human artists to generate new works.
Beyond the issue of copyright infringement, AI-generated art raises critical questions regarding the very nature of creativity. While AI is capable of producing visually compelling images, it lacks the human intuition, emotion, and personal experience that underpin genuine artistic expression. Critics argue that AI-generated art, despite its technical prowess, is deficient in the authenticity and depth intrinsic to human creativity. This concern accentuates the broader ethical debate about the role of AI in the arts and the potential diminution of the uniquely human elements of creativity. This ethical debate necessitates a more sophisticated approach to incorporating AI in the creative industries, one that rigorously considers the rights of human creators and the long-term implications for the future of artistic expression. Such an approach must balance the innovative potential of AI with the preservation of human creativity as the cornerstone of artistic and cultural production.
Balancing Innovation and Protection: A Way Forward
In recognition of the complexities surrounding AI-generated works, there have been increasing calls to amend the Indian Copyright Act to accommodate these technological advancements. One proposal is to categorise AI as a tool rather than an author, ensuring that ownership of AI-generated works resides with a natural or juristic person. Such legislative changes would provide much-needed clarity and accountability while adapting the legal framework to the realities of technological innovation.
The Indian government has proactively initiated measures to harness the potential of AI through policies such as the 'AI for All' initiative and the establishment of the AI Task Force, which focuses on leveraging AI for social and economic development. Nevertheless, these initiatives are undermined by the absence of a comprehensive legal framework that adequately addresses the unique challenges introduced by AI-generated content. In 2021, acknowledging the growing significance of AI across diverse sectors and its prospective economic ramifications, the Parliamentary Standing Committee Report advocated for a thorough revision of intellectual property rights legislation. The Report emphasises the necessity of constructing a robust IPR framework to "extract benefits from AI," citing research from Accenture that projects AI-related advancements could substantially bolster the Indian economy by 2035. Despite providing a broad strategic vision, the Report falls short of delineating specific issues or offering concrete, actionable AI and intellectual property recommendations.
As AI-generated art continues to expand the horizons of creative expression, it becomes increasingly imperative to balance the promotion of innovation with the protection of intellectual property. Navigating this intricate landscape necessitates the implementation of several strategic measures. Firstly, there is an urgent need to establish more precise guidelines governing AI and copyright. The current Indian Copyright Act does not recognise AI as an author of creative works, thereby creating a significant lacuna in the legal framework as AI's influence on creative production intensifies. This deficiency necessitates a critical re-evaluation of existing legislative provisions and their applicability to AI-generated works. Given the swift advancements in AI technology, legislative amendments are crucial to address the recognition of AI's role in authorship and to mitigate the myriad legal issues arising from AI-generated content.
Secondly, safeguarding the interests of small artists and creators is paramount. The law should incorporate mechanisms that enable these individuals to protect their works from being utilised in AI training datasets without explicit consent. This could involve mandating greater transparency from AI companies regarding the data they employ and instituting more stringent consent requirements to ensure that artists retain control over their intellectual property. Thirdly, the doctrine of fair use should be expanded and refined to better align with the realities of the AI era, wherein algorithms may generate derivative or transformative works. Clear and precise guidelines are necessary to delineate what constitutes fair use of existing copyrighted material by AI, thereby ensuring a balanced approach that respects the rights of original creators while fostering technological innovation.
Moreover, moral rights, such as the right to integrity, must be adapted to apply within AI contexts. New legal frameworks should ensure original creators are duly recognised and their works respected, even when AI tools transform them. This adaptation is essential to maintain the relevance and efficacy of moral rights in an evolving technological landscape, thereby safeguarding the personal and reputational interests of human creators. Furthermore, implementing stricter data regulations is imperative to address ethical concerns related to data usage. A robust data privacy framework would protect individual rights while simultaneously promoting innovation, ensuring that AI companies do not infringe upon the copyrights of artists. Such regulations would provide a clear legal basis for addressing grievances related to the unauthorised use of creative works in AI training processes.
AI-generated art transcends mere technological novelty; it represents a fundamental shift in our conceptualisation of creativity, ownership, and authorship. While AI furnishes limitless possibilities for artistic and creative endeavours, it simultaneously challenges entrenched legal and ethical norms, particularly in intellectual property. The Indian legal system, deeply rooted in the principles of human authorship, faces a formidable challenge in adapting to these transformative changes. As AI technology continues to evolve, it is imperative for India to undertake a comprehensive reassessment of its intellectual property framework and to promulgate new legislation that judiciously balances the interests of AI developers, artists, and society at large. By fostering a legal environment that encourages innovation and protects the rights of creators, India can ensure that AI-generated art serves as a catalyst for creativity rather than a source of legal and ethical conflict.
[1] See China, Case Reference (2023) Jing 0491 Min Chu No. 11279] (27 November 2023): A landmark decision where copyright was upheld in original works, distinguishing between human-authored content and AI-generated material. And S. Š v. TAUBEL LEGAL (Municipal Court in Prague, No. 10 C 13/2023-16): this case concluded that DALL-E, as an AI, cannot be recognised as an author of a work, reinforcing the principle that authorship must be attributed to a human.
[2] Getty Images v Stability AI (UK): The potential significance of the Stability AI case, which is currently pending trial before the High Court in London, lies in the defence that Stability AI intends to raise against Getty Images' claims. The outcome of this case could set a precedent in how copyright law is applied to AI-generated content and the use of copyrighted material in training AI models. There has been a significant focus on the potential for creating a new copyright framework specifically designed to address the challenges posed by AI training.
*Professor of Law and Technology, School of Law, University of Leeds
**Independent Scholar