AI Content and Copyright: A Criticial Analysis

The Author, Nazifa Khateeb, is a second-year law student at Maharashtra National Law University, Mumbai. She is currently interning with LatestLaws.com and the Indian Dispute Resolution Centre.

The rapid spread of generative Artificial Intelligence (AI) has presented challenge to the foundational principles of intellectual property law globally. In India, this challenge manifests as a direct conflict between the human-centric paradigm of the Copyright Act, 1957, and the increasingly autonomous nature of AI-generated creative works.[1] The core of the issue lies two distinct but interconnected problems: the copyrightability of AI-generated outputs and the legality of using copyrighted materials to train AI models. On the first front, India's copyright law, particularly the definition of "author,"[2] is predicated on human creativity, skill, and judgment. While a 1994 amendment introduced the concept of an author for "computer-generated" works, its application to modern generative AI remains fraught with ambiguity. Administrative decisions from the Indian Copyright Office, mirroring global trends, have affirmed a "bedrock requirement of human authorship," denying protection to works where AI is the primary creator and leaving the crucial question of what constitutes "sufficient human input" unanswered.[3]

On the second front, the very process of AI training, which involves the mass reproduction and storage of copyrighted data, constitutes a prima facie act of infringement under the Act. AI developers' primary defense rests on the "fair dealing" exceptions in Section 52,[4] a narrow and enumerated set of provisions ill-suited to the context of commercial text and data mining. This has forced a jurisprudential debate, now playing out in the Delhi High Court, on whether to import purposive doctrines like "transformative use" to create a judicial safe harbor for AI training.

This legal uncertainty is compounded by a notable policy impasse. While a Parliamentary Standing Committee has recommended a comprehensive legislative overhaul, including the creation of a new category of rights for AI, the executive government has publicly maintained that the existing legal framework is "well-equipped" to handle these challenges. This stance effectively outsources the initial, difficult policy-making to the judiciary, placing immense pressure on the courts to resolve fundamental questions of national innovation strategy through case-by-case adjudication.

The landmark lawsuit, ANI Media v. OpenAI,[5] has become the crucible for these issues. The case did not only test the limits of "fair dealing" but also forced the judiciary to choose between a creator-centric model that prioritizes existing rights holders and an innovator-centric model that facilitates technological development. The outcome, along with a comparative analysis of approaches in the United States, United Kingdom, and European Union, will be pivotal.

India stands at a critical juncture, and the path it chooses, whether through nuanced judicial evolution, targeted legislative intervention, or a hybrid of both, will have lasting consequences for its ability to balance the protection of creativity with the promotion of a globally competitive AI industry.

copyrights core under ai strain

The Copyright Act, 1957,[6] as amended over the decades, forms the bedrock of intellectual property protection for creative works in India. Enacted long before the advent of modern artificial intelligence, its provisions are now being stress-tested by the unique challenges posed by generative AI.

1. idea expression dichotomy

Indian copyright law is built upon several fundamental principles that are now being re-examined in the context of AI. The first is the idea-expression dichotomy, a universally recognized concept which posits that copyright protects the unique expression of an idea, but not the idea, fact, or concept itself. For example, the idea of a tragic romance between members of rival families is not protectable, but a specific play or novel expressing that idea is. This distinction is critical for AI, as developers often argue they are training models on unprotectable facts, data, and linguistic patterns, while rights holders contend that this process involves the copying of protected expression.

2. the originality standard

The second core principle is originality. Section 13[7] of the Act grants copyright to "original literary, dramatic, musical and artistic works". While the Act does not define "original," Indian jurisprudence has moved away from the low-threshold "sweat of the brow"[8] doctrine (which grants copyright based on mere labor and diligence) towards a higher standard requiring a modicum of "skill, labour, and judgment"[9] from the author. The central question is whether AI-generated content, produced by algorithms processing vast datasets, can meet this standard. Opponents argue that AI lacks the requisite human consciousness and judgment for true originality, merely recombining existing patterns in a sophisticated manner.

3. human centric paradigm

This leads to the third and most pervasive principle: the human-centric paradigm. The entire architecture of the Copyright Act, 1957,[10] presupposes a human creator. Copyright is designed to provide incentives to human authors, protecting their creations and rewarding their creativity. The concept of moral rights (paternity and integrity), the term of copyright being tied to the author's lifespan, and the very definition of "author" are all deeply rooted in the notion of human intellect and effort. Generative AI, capable of producing complex works with minimal or ambiguous human intervention, fundamentally disrupts this paradigm, forcing a re-evaluation of what it means to "create" and who or what can be an "author".

the question of authorship – the core friction

The legal friction between AI and copyright law is most apparent in the Act's core definitions. Section 2(y) of the Copyright Act[11] defines "work" to include literary, artistic, musical works, films, and sound recordings. This is broad enough to cover AI-generated content like text, images, or music. The real legal tension lies in the concept of "authorship" under Section 2(d),[12] which assigns authorship to human creators (e.g., writers, artists, composers). This humanist framework excludes AI as a legal author. Section 17[13] is what designated this by stating that the "author" is the first owner of copyright. Since AI cannot qualify as an author, it challenges the very existence of copyright in AI-generated works. While exceptions like the “work-for-hire” doctrine may help developers or users claim ownership, such claims are moot unless a legally recognized author exists in the first place.

The most debated provision in this entire landscape is Section 2(d)(vi)[14] of the Copyright Act. Introduced by the Copyright (Amendment) Act, 1994, it defines the author "in relation to any literary, dramatic, musical or artistic work which is computer-generated" as "the person who causes the work to be created".

This provision is a double-edged sword. On one hand, it demonstrates remarkable foresight by the Indian Parliament, creating a specific category for computer-generated works and acknowledging that they can have an author, a step many jurisdictions did not take at the time. This seemingly provides a direct statutory pathway to grant copyright to AI-generated content.

On the other hand, the provision is the primary source of legal ambiguity. It was drafted in an era when "computer-generated" likely referred to works created by software where human input had a direct, predictable, and causal effect on the output. The phrase "the person who causes the work to be created" fit this model well. However, with modern generative AI, the causal link is blurred. A user provides a prompt which can also be called the idea, but the AI model, drawing on its vast training and complex neural networks, generates the expressive output in a process that is often unpredictable. This raises the main challenge of: who is the "person" that "causes" the creation?

The provision's wording presupposes a singular "person" and a clear act of "causing," concepts that break down when faced with the distributed and collaborative nature of AI creation. The very legislative tool designed to provide clarity for technology has become the central battleground for legal interpretation in the age of AI. The debate has shifted from whether a computer-generated work can have an author to defining the threshold of human involvement required to be considered the "causer" of that work.

If selecting inputs is insufficient, what level of human involvement is sufficient? This leads to the debate over "prompt engineering." Proponents of copyright for AI-generated works argue that crafting a detailed, specific, and iterative series of prompts is a creative act that involves significant skill and judgment. They contend that a complex prompt is not just an idea but a form of expression that guides the AI's output in a controlled manner, akin to a director giving detailed instructions to an actor. However, this view is also contested. The U.S. Copyright Office, in its influential guidance, has taken the position that prompts are generally unprotectable "ideas" that do not give the user sufficient "creative control" over the final work. The AI model, not the user, ultimately determines the final expressive elements. The prompt "a high-quality photograph of an astronaut riding a horse" is an instruction, but the AI generates the specific composition, colors, and details that constitute the copyrightable expression.

A potential middle ground may exist. Indian law could evolve to recognize a distinction between a simple, one-line prompt and a sophisticated process of prompt engineering involving dozens of iterations, negative prompts, and parameter adjustments. If a user can demonstrate that their detailed inputs and iterative refinement process effectively "caused" the final work to take its specific expressive form, they might have a stronger claim to authorship under Section 2(d)(vi).[15] However, as of now, this remains a theoretical argument without clear judicial or legislative endorsement in India.

the economic & ethical divide

Several compelling arguments support extending copyright protection to AI-generated works. Incentivizing innovation and investment is a key justification. The development of sophisticated AI models requires enormous capital investment. If the outputs of these models cannot be protected by copyright, developers may struggle to monetize their creations, potentially stifling innovation. Similarly, users and creative professionals will be less likely to invest time and resources in learning to use these tools if the resulting works are immediately cast into the public domain, unable to be licensed or sold exclusively. For India to position itself as a global hub for AI development, as envisioned by government reports, a robust IPR framework that provides economic rewards is seen as essential.[16]

Proponents argue that copyright should reward the human effort involved in the creative process. Even if the AI does the "heavy lifting," the user often invests significant time, skill, and creative vision in conceptualizing the work, refining the prompts, and curating the outputs. This aligns with the rationale behind the "sweat of the brow" doctrine, which, although not the prevailing standard in India, reflects an underlying principle of rewarding labor. The user, as "the person who causes the work to be created," is the most logical candidate for authorship under Section 2(d)(vi).

More radical argument proposes granting legal personhood to AI. If an AI could be recognized as an author, it could own copyright in its creations. This would, in theory, foster a competitive creative landscape where AI-generated works compete alongside human works, potentially spurring greater innovation from both. This approach, however, would require a fundamental overhaul of the legal system and is not currently considered a viable option in any major jurisdiction.

However, equally if not more powerful are the arguments against granting copyright to AI-generated works. The primary objection is a lack of originality and creativity. Copyright law protects works that are the product of an author's skill, judgment, and intellectual effort. Additionally, AI being a non-sentient machine, cannot possess these human qualities. Its outputs are sophisticated derivations and recombination of the data it was trained on, not truly original creations born from an independent intellect. As there is no "author" in the human sense, there can logically be no copyright.

Secondly, granting copyright to AI would be inconsistent with the fundamental purpose of copyright law, which is to incentivize human creativity. Machines do not require legal or economic incentives to create; they simply execute their programming. Extending copyright protection to them would not further the constitutional goal of promoting progress in the arts and sciences.

Thirdly, there are significant concerns about market disruption and the devaluation of human creativity. If high-quality creative works can be generated nearly instantaneously by AI and receive copyright protection, it could flood the market, making it difficult for human artists, writers, and musicians to compete and earn a livelihood. This could disincentivize human creation, perversely undermining the very goal of copyright.

And lastly, there are intractable technical and policy problems, most notably the term of copyright. In India, copyright for literary and artistic works typically lasts for the life of the author plus 60 years. If an AI were considered an author, it would be effectively immortal, leading to a perpetual copyright. This would prevent works from ever entering the public domain, which is a crucial part of the copyright lifecycle that allows for future creativity and access to cultural heritage. In light of these issues, many argue that the most logical and socially beneficial outcome is for all works generated purely by AI to enter the public domain immediately.

The question of who owns copyright in AI-generated works, assuming such works are protectable, has given rise to multiple models. One view favors the user (prompter) as the author, since they initiate the creation through prompts, an argument supported by some AI platforms’ terms of service, though limited by statutory interpretation. Another model supports the developer, claiming ownership under work-for-hire principles. A joint ownership model suggests both user and developer share rights, but this raises practical issues around control and royalties. The idea of the AI itself owning copyright is currently untenable under Indian law due to its lack of legal personhood. If no human authorship can be established, the work likely falls into the public domain. India’s current legal position rejects copyright in works created by fully autonomous AI or with minimal human input, signaling that future claims will depend on demonstrating a threshold of significant human creative contribution.

judicial intervention & nuances

In the landmark case of ANI v OpenAI,[17] news agency ANI filed a suit against OpenAI, alleging that the unauthorized use of its journalistic content, including paywalled articles, images, and videos to train ChatGPT amounts to copyright infringement. ANI claims this violates its rights under the Copyright Act, particularly those of reproduction and adaptation, and further alleges reputational harm due to ChatGPT generating false information wrongly attributed to the agency. OpenAI, in its defense, challenged the Delhi High Court’s jurisdiction by asserting that training occurred on servers outside India. Substantively, it invokes the fair dealing exception under Section 52,[18] arguing that its models process facts and linguistic patterns not protected expression and thus do not infringe ANI’s rights. The Delhi High Court, recognizing the broader implications, has framed four key legal questions. Firstly, whether the storage and use of ANI’s content during training and output constitute infringement? Secondly, if fair dealing applies; and whether the court has jurisdiction. Two amici curiae offered contrasting views. One argued for a broad, innovation-friendly interpretation, likening machine learning to human reading and calling for a purposive reading of the research exception. In contrast another took a strict textualist view, arguing that mere storage constitutes infringement and that India’s fair dealing law is narrowly defined and exhaustive. The case gained further significance when major music labels like T-Series and Saregama intervened, alleging similar unauthorized use of their content. This expansion elevates the case from a dispute over news data to a precedent-setting battle that will shape the future of how AI engages with all copyrighted content in India.

Furthermore, though not a copyright case, another recent Delhi High Court decision provides crucial context for the judiciary's approach to AI. In Anil Kapoor v. Simply Life India & Ors.,[19] the court granted a sweeping injunction protecting the famous actor's personality rights against unauthorized use by generative AI tools. The court restrained defendants from using his name, image, voice, and other attributes for commercial purposes without permission, including through AI-generated deepfakes and morphed images.

This case is significant because it demonstrates the Indian judiciary's willingness to proactively adapt existing legal principles, in this instance, the common law right of publicity or personality to address the novel threats posed by AI. Even without a specific statute governing AI-generated fakes, the court recognized the potential for harm and fashioned a remedy. This pro-creator and pro-rights holder stance signals that the courts are not hesitant to intervene to protect individuals and their creative personas from technological misuse. This judicial attitude could well influence the court's perspective in the ANI copyright case, suggesting a potential inclination to protect the rights of content creators against the unauthorized exploitation of their work by large technology platforms.

India's struggle with AI and copyright is not occurring in a vacuum. Major economic and legal jurisdictions around the world are facing the same challenges, and their diverse approaches provide a valuable comparative lens through which to assess India's potential paths forward. A global consensus is emerging on some issues, while on others, fragmentation and experimentation are the norms.

The United States has been a key battleground for AI-copyright disputes, and its jurisprudence has been highly influential in India. The most significant development on the authorship front is the case of Thaler v. Perlmutter.[20] Here, the U.S. Court of Appeals for the D.C. Circuit affirmed the Copyright Office's refusal to register a work autonomously created by an AI, holding that U.S. copyright law "requires all eligible work to be authored in the first instance by a human being". This establishment of a "bedrock requirement" of human authorship has been explicitly cited and echoed by the Indian Copyright Office in its Ankit Sahni[21] decision, indicating a strong jurisprudential alignment on this point.

Where the U.S. approach diverges significantly is on the issue of AI training. The legality of using copyrighted works to train AI models is being fiercely litigated in numerous high-profile cases, such as The New York Times Co. v. Microsoft Corp. & OpenAI[22] and Andersen v. Stability AI.[23] The central defense in these cases is the doctrine of "fair use". Unlike India's rigid "fair dealing," U.S. fair use is a flexible, equitable doctrine that courts apply on a case-by-case basis using a four-factor test: (1) the purpose and character of the use; (2) the nature of the copyrighted work; (3) the amount and substantiality of the portion used; and (4) the effect on the potential market. AI companies argue their use is "transformative," a key consideration under the first factor, because they are creating a new tool rather than a substitute for the original works. This flexible, case-driven approach provides a stark contrast to India's more restrictive statutory framework.

The United Kingdom's approach offers a potential legislative middle path. The UK government has been actively exploring reforms to its copyright law to address AI. One of the most prominent proposals is the creation of a broad Text and Data Mining (TDM) exception to copyright infringement. This exception would permit the use of copyrighted works for training AI models for any purpose, including commercial ones, under one crucial condition: that the rights holder has not explicitly "opted out" of such use.

This model attempts to balance the needs of innovators and creators. It would give AI developers a clear legal pathway to access training data while empowering rights holders with the ability to control the use of their works by reserving their rights, likely through machine-readable signals. Interestingly, the UK is also contemplating removing its existing copyright protection for purely computer-generated works a provision that currently makes UK law more permissive than that of the US or EU. This "opt-out" system represents a clear, rule-based legislative solution that India could consider as an alternative to relying solely on judicial interpretation of its outdated "fair dealing" provision.

The European Union has taken a distinct, regulation-centric approach with its landmark AI Act. While the Act is a comprehensive framework for AI governance and not a copyright law per se, it contains crucial provisions that directly impact the copyright debate.

A key feature of the EU AI Act is its focus on transparency. It imposes obligations on developers of "foundation models" (like GPT-4) to maintain detailed documentation of their training processes. Most importantly, it requires them to "draw up and make publicly available a sufficiently detailed summary of the content used for training" the model.

This transparency requirement does not resolve the question of whether the training itself is copyright infringement. However, it provides a powerful tool for rights holders. By compelling disclosure, the law ensures that authors, artists, and publishers can discover whether their works have been used to train a particular AI model, which is a critical first step in enforcing their rights. This regulatory approach, focused on accountability and information symmetry, offers another policy tool that India could adopt, complementing any potential changes to its copyright law.

The global landscape reveals a clear divergence in approaches. On the question of authorship, a strong consensus is forming around the U.S. position, requiring human creativity as a prerequisite for copyright protection. India appears to be firmly in this camp.

However, on the more contentious question of AI training, the paths are fragmented. The U.S. is relying on its flexible, judge-led "fair use" doctrine to resolve disputes. The UK is actively considering a specific, legislated "opt-out" exception for TDM. The EU is using its regulatory power to enforce transparency.

India currently finds itself in a uniquely challenging position. It has adopted the restrictive American view on authorship but lacks the flexible American defense for training. Its domestic law has neither the specific legislative safe harbor being considered by the UK nor the robust regulatory transparency mandate of the EU. This leaves India's "fair dealing" doctrine, a rigid and narrow provision, as the only and likely inadequate tool to address one of the most complex technological and legal challenges of our time. This international context underscores the immense pressure on the Indian judiciary in cases like ANI v. OpenAI and highlights the urgent need for a clear policy direction.

Conclusion

India stands at a pivotal moment in the development of its intellectual property jurisprudence. The collision of generative AI with the Copyright Act, 1957, has exposed the limitations of a legal framework designed for a different technological age. The core tenets of copyright—authorship, originality, and the rights of reproduction—are being fundamentally questioned. The resulting legal ambiguity, characterized by the dual challenges of output ownership and training data legality, has created a high-stakes environment for both India's burgeoning technology sector and its vibrant creative industries.

The current policy of strategic ambiguity, where the government publicly defers to the courts while privately studying the need for reform, is a temporary and unsustainable solution. While it allows the judiciary to perform the initial, difficult work of legal pathfinding, it leaves innovators and creators in a prolonged state of uncertainty. The ANI v. OpenAI case has become the focal point of this national debate, with the Delhi High Court tasked with making a decision that will have far-reaching policy implications.

Ultimately, navigating this new frontier will require a move beyond the current impasse. The most viable path forward for India is likely a hybrid one: allowing the courts to establish foundational principles through reasoned interpretation of existing law, followed by swift and targeted legislative action to codify these principles and address the specific gaps that judicial interpretation cannot fill. By embracing a balanced approach that fosters innovation through clarity while respecting and rewarding human creativity, India can build an IPR regime that is not only fit for the age of AI but also serves as a model for the developing world. Coherent, forward-looking, and decisive action is no longer an option, but an imperative.

References:

Picture Source :