Recently, there has been some news about the re-release of the film Raanjhanaa, starring Dhanush and Sonam Kapoor. The director, however, has objected to the release of this version. Tejaswini, a very enterprising scholar, has written a detailed post on SpicyIP unpacking the issue. During a conversation with her, I shared a text offering my two cents on the matter. I’m reproducing that text below—with a few edits to make it a more readable version. I must say that I’ve been interested in this issue for quite some time, and had even written about it for the first edition of SpicyIP‘s Shamnad Basheer Essay Competition, where my entry on this very topic was awarded first place. See also here and here.
Okay. Here’s what I wrote to her…
“In this case, I think the director’s moral rights claim is weak—perhaps even a non-issue, both legally and conceptually.
Why?
The director wants to dissociate from the film because the producer is changing its ending, allegedly altering its meaning. But there’s no specific claim of reputational harm. Nor is there a dispute over attribution. So, which moral right is actually being invoked here?
Under Indian copyright law, Section 57 gives two main moral rights:
The Right of Attribution – which the director isn’t asserting.
The Right of Integrity – which protects against distortion, mutilation, or modification that harms the author’s honour or reputation.
But here, the director’s objection rests on something vaguer: discomfort over a perceived shift in the film’s meaning. That alone doesn’t amount to reputational harm. There’s no apparent injury to honour or dignity—just disapproval of an interpretive direction.
To me, ’tis a philosophical objection, reminding me of Roland Barthes‘s famous article called the Death of the Author, where he argues that the meaning isn’t in the author. Once a work is public, its meaning/interpretation is no longer controlled by its creator.
This also brings to mind Abhay Deol’s reading of the film, where he expressed discontent with the movie’s message. Yes, people often speak of “meanings” in art and cinema, but I wonder what they mean by “meaning.” No single review score or Rotten Tomatoes rating can convey the meaning of a film. It’s because there is no singular meaning built into the movie. Viewers extract different meanings, often contradictory ones, and yet respond similarly.
The myth of a singular, stable meaning must be busted.
And even if the ending is altered, that’s not per se wrongful. Gestalt theory is also an interesting way to look at it, which the Delhi High Court in MRF tires also reinforced, that the whole is greater than the sum of its parts—a new ending simply reorganises meaning.
Maybe viewers will now see a one-sided lover who “gets the girl” who once tried to kill him. Is it a happy ending? Maybe for you. Not for me, necessarily.
If not moral rights, what’s the actual issue here? Perhaps … it’s the question of control: Can a contributor—who may not even be the author—prevent the rights-holder from altering the work’s meaning?
That’s where things get interesting.
Under Indian copyright law, moral rights don’t go that far. Economic rights might, if the director is a co-author with a say over derivative works. But most likely, he isn’t.
But herein lies a hitch: Section 2(d) of the Copyright Act, 1957 doesn’t define “author” as such—it instead merely assigns the title and tells us who the author is. Put otherwise, it does not say what makes someone an author.
This matters.
If authorship were based on creativity or contribution, directors might qualify. But Indian law prioritises control and investment. Especially for the producers who are the authors of the cinematography work, the law concern isn’t creativity—it’s capital. It recognises the one who pays, not necessarily the one who creates.
So, if the director has no authorship/ownership stake and contractual arrangement, he’s out of luck.
If you’re interested in exploring this topic further, you may want to look into Auteur theory. Historically, the question of the director’s creative authority has surfaced at least twice—once during the 1967 Revision Conference of the Berne Convention, and later in the context of the 2010 Amendment Bill in India. I have explored the issue in depth here in this piece. Director’s Authorship under Indian Copyright Law: An (Un)Indian Approach? (January 18, 2021). Journal of IP Studies, NLU Jodhpur, Available at SSRN: https://ssrn.com/abstract=3768248
A Few days ago, Akshat wrote a piquing post problematising the MEITY report, which concludes that training Large Language Models infringes and isn’t protected under Section 52(1)(a)(i) of the Copyright Act.
In his casually complex style, he noted:
“What’s missing from the current debate is a crucial understanding: copyright protects against unauthorized sharing of my work with others, potentially depriving me of credit or economic compensation that I could have gotten by sharing it with them myself. In simpler terms, while I cannot express your original expression without your permission, I can certainly consume your publicly available original expression without the same (maybe (or not?), barring paywall circumvention, which isn’t part of this current debate). The law focuses on unauthorized expression of original publicly available content—not its unauthorized consumption, as making content public already waives that claim.”
Reading Akshat’s piece and reflecting on my previous works, I don’t think there’s any angle that is innocently missed. The people who wrote the MEITY report very likely knew (and should have known!) that AI training can involve a different kind of “use” compared to the type of usage the Copyright Act authorises owners to exclude. Not every use constitutes infringement—only expressive ones, as Aksaht aptly pointed out. Not every act of storing a work is problematic, either. If no human involvement or communication occurs, why apply the fair use doctrine? Fair dealing, I repeat, is a tricky terrain, and sometimes it’s best not to tread there.
But I sense something more at play here, something beyond what we consider good or bad arguments, or what makes for a desirable or undesirable policy. From my limited readings, I’ve observed the AI and copyright discourse, and it feels like a loop—repackaged arguments, recycled citations, well-worn tropes of author vs. public, incentive vs. access, and innovation vs. regulation. While there are some shifts in interpretation—debates now incorporating the technicalities and workings of AI, discussing whether a use qualifies as fair, its alignment with copyright’s purpose, or invoking the balance trope—few changes reshape the discourse in any meaningful way.
This makes me wonder: Has the discourse around AI-Copyright, the way we think about the issue, been set, controlled and regulated? As Foucault says, “In every society the production of discourse is at once controlled, selected, organised and redistributed by a certain number of procedures whose role is to ward off its powers and dangers, to gain mastery over its chance events, to evade its ponderous, formidable materiality.”
Who knows and how: whether we remain caught in the act of defending or extending our current positions? I attempted to address this question in my upcoming piece for the Journal of the Copyright Society of the USA, where I argue that a “balance” discourse is grounded in a utilitarian episteme—the general system of thought—that shapes the current approach to the issue. However, I feel I have left the issue incomplete. There is a deeper layer here that extends beyond the existence of a discourse.
It is also a question of agency: who controls the discourse itself, how the boundaries of discourse are defined, and who has the authority to do so. There are “epistemic forces” at play which shape not just what is said, but how it’s said and by whom. The questions we ask, the terms we use, and even the arguments we entertain are all filtered through an underlying power/struggle. What I’ve missed, and what I believe Akshat also missed in his post, is that this discourse delimitation is not passive or neutral. It is actively shaped by those who have the power to define the terms of the debate—and by those who have the power to exclude alternative narratives or conceptions from entering the conversation.
So, why assume that an argument or conceptual angle is absent simply because people are unaware of it or fail to appreciate its significance? It could be a deliberate omission. The way copyright and AI discourse is unfolding—especially in the Global North, and now in India and China—isn’t simply a matter of organic, unregulated exchange. The way courts and policymakers approach the issue, and how it’s shaped by the agendas of those in positions of power, is far from incidental. This discourse is actively moulded, steered, and controlled by the dominant players in the field. It’s not a free-flowing debate where the most compelling and convincing claims naturally prevail. No, it’s carefully shaped and restricted by the agents involved—not just by the substance of what’s being said, but by who gets to speak, and more crucially, by the language in which those arguments are framed.
Akshat has highlighted Big Tech’s influence through ChatGPT’s Opt-Out mechanism at SpicyIP, another blog for which both of us write. Similarly, we also know that tech companies are lobbying for easier access to copyrighted works. Meanwhile, authors and rights holders are pushing for licensing or remuneration. The Standing Committee on Copyright and Related Rights of WIPO is drafting a proposal on author remuneration, a key agenda item at the upcoming March 2025 meeting. Civil society groups and scholars are advocating for new research exceptions. Even those who argue that AI training doesn’t constitute infringement still (need to) frame their analysis in terms of fair dealing as a safeguard because that’s how legal arguments are structured in courtrooms and policy debates. The field is structured to work in this way.
On top of this, since the 1980s, we’ve arguably gestated a faith in fair dealing as the default response to every new technology. It has become a kind of conceptual messiah—the go-to solution, regardless of context. At the same time, authors and copyright holders know that some uses of their works are inevitable. They can’t stop them, so they focus on monetisation—hence the push for licensing as a practical compromise.
The real issue isn’t just what arguments are being made, but who is making them, where, and why. This is a symbolic struggle for power, with different groups competing to define the world order and how copyright law fits into it. In turn, there is an effort to define how knowledge governance will happen in the coming years.
That’s all from my end for now. I’ve explored these questions in greater depth in an upcoming piece for the Indian Journal of Law and Technology, co-authored with Luca Schirru, a brilliant scholar from Brazil and a dear friend. I will share that piece once it is published.
Today is World IP Day – that annual celebration where global institutions glorify a system supposedly designed to “encourage innovation” and “reward creativity.” As much as I’d love to join the festivities, I find myself unable to raise a glass to a system that has betrayed its original purpose.
Let me be blunt: modern intellectual property regimes have morphed from shields meant to protect creators from market pressures into weapons that deepen their market dependency and actively distort cultural production. The recent Daagarvani case in India, where a court granted relief against a song composed in the same traditional Ragas as another, demonstrates how copyright’s expansionary logic now threatens centuries-old cultural practices and practitioners. Meanwhile, the rise of AI systems trained on vast datasets of human creativity intensifies these contradictions potentially, through arguments by creators and publishers facing existential crises, by commodifying creative capacity itself rather than merely creative works.
But these examples are merely symptoms of a deeper theoretical problem – one that demands we reconsider the fundamental nature and purpose of intellectual property itself. What if IP’s current form isn’t an inevitable or natural development, but rather a historically contingent arrangement that has become unmoored from its original protective purpose?
Let’s get something straight about IP’s origins. Those early regimes – from the Venetian printing privilege of 1469 to England’s Statute of Anne in 1710 – weren’t designed to “incentivize” creation through market rewards. Their actual purpose was to protect creators and disseminators from emerging market pressures so they wouldn’t be discouraged from producing socially valuable expression. It was to enable them and for their economic security in a market society (not luxurious accumulation)
The Statute of Anne explicitly speaks of preventing the economic “ruin” of authors and their families – not maximizing creative output or maximal “monetizing” at the cost of access to such intangible resources. Early patent systems emphasized protection for investments already made, not incentives for future innovation. These systems were fundamentally about providing basic economic security for cultural producers in increasingly competitive market environments.
But here’s where things went sideways: by conceptualizing this protection in terms of market-compatible property rights rather than direct provision or alternative support systems, these regimes inadvertently laid the groundwork for the commodification of cultural expression and information itself. What began as attempts to shield creators from market pressures gradually transformed into mechanisms for deeper market integration and expansion.
This transformation stemmed from several theoretical misconceptions. First, a fundamental error similar to what Dianne Elson identifies as adhering to a “labor theory of value” rather than developing a “value theory of labor” – naturalizing the relationship between labor and value, creating the illusion that market value is attributable to intrinsic qualities of creative works rather than to specific social relations that impact labour used to produce such output. Second, the “pre-social creativity” myth that ignores the fundamentally social and communicative nature of cultural production. Third, “physicalism” that inappropriately attempts to derive normative conclusions about nonrival informational resources from physical descriptions of property relations.
Over time, intellectual property’s conceptual framing shifted from protection against market pressures to “encouragement” through market incentives. Its institutional form increasingly converged with conventional property paradigms despite fundamental differences between informational resources and their non-rivalrousness and physical goods. Its scope and duration dramatically expanded through legislative amendments and judicial interpretations, and its relationship to creative practice became increasingly distorted.
This market-based structure generates several distinctive distortionary effects that undermine IP’s purported goals. Let me walk you through three crucial ones:
By creating artificial scarcity for inherently non-rival resources, IP enables pricing information goods above their marginal distribution cost (which for digital goods approaches zero). This pricing structure excludes potential users whose ability to pay falls below the market price but above the actual cost of providing access.
Beyond efficiency concerns, this raises profound questions of distributive justice and cultural participation. When access to informational resources depends primarily on willingness and ability to pay, distribution inevitably skews toward those with greater financial resources regardless of creative potential or social contribution.
Market-based allocation through IP amplifies preferences of economically advantaged groups while marginalizing others. Because markets respond to willingness and ability to pay rather than need or potential social contribution, investment naturally flows toward satisfying demands of affluent consumers while neglecting less economically powerful communities.
This manifests across various domains – from pharmaceutical research prioritizing wealthy-world conditions over “neglected diseases” affecting billions, to entertainment industries concentrating investment in products targeting demographics with greatest disposable income while culturally significant but less commercially viable expressions receive minimal support.
Perhaps most troublingly, IP systems create systematic biases regarding which types of creative activities receive investment and development. This “distortionary effect” operates through market mechanisms that value certain characteristics of informational goods over others, independent of their intrinsic merit or social contribution.
IP’s reliance on excludability – the capacity to control access and use – as the primary mechanism for monetizing informational goods creates inherent preferences for innovations and creative works that exhibit high excludability and appropriability through existing property rights, regardless of their social value compared to less excludable alternatives.
Before proceeding further, we should recognize that defenders of strong intellectual property rights have marshaled various theoretical justifications for exclusive control. These fall into several main traditions:
Labor-desert theories, often traced to John Locke, suggest creators deserve compensation for resources they produce through their labor. However, with nonrival resources, labor theories at most justify claims to fair compensation, not exclusive control. When someone creates information, preventing others from using it cannot be justified as protecting the creator’s own use (which remains unaffected by others’ use) but solely as protecting compensation interest.
Personality theories, associated with continental European traditions, emphasize the intimate connection between creators and their works. Yet the personality interests these theories identify – primarily concerning recognition of authorship and integrity – do not inherently require restricting others from using creative works. These interests can be protected through attribution and integrity rights without granting comprehensive control.
Utilitarian theories justify IP through its role in addressing market failures in information production. However, this account provides no intrinsic justification for exclusive control over information – only adequate compensation to maintain production incentives. This compensation interest doesn’t inherently require exclusive rights but merely sufficient revenue to recover risk-adjusted development costs plus normal returns.
Democratic theories evaluate IP by its contributions to robust democratic culture and equitable access to conditions of self-determination. These approaches typically support limitations on exclusivity to promote widespread participation in cultural meaning-making.
As Profs. Talha Syed and Oren Bracha remark, none of these traditions provides compelling first-order justification for exclusive control rather than appropriate compensation or recognition. The strongest remaining defense lies in second-order institutional considerations – IP’s role in harnessing decentralized market signals of information to guide innovation and prices. Yet this justification remains parasitic on first-order values and contains inherent expansionary tendencies.
The Indian classical music case of Ustad Wasifuddin Dagar v. A.R.Rahman and Ors. exemplifies IP’s inherent expansionary logic. The court granted relief against a song sharing ragas with another composition, ignoring that such similarity is both inevitable and desirable within this tradition. Every raaga has rules making perceivable similarity between compositions inevitable, including characteristic phrases called “pakad” that signify the raaga’s identity.
This case isn’t an isolated incident but part of a systematic pattern where exclusionary rights expand beyond justifiable boundaries to encompass elements that should remain in the cultural commons. The same dynamics appear in hip-hop sampling, documentary filmmaking, fan fiction, and countless other creative practices that necessarily build upon existing materials.
Similarly, the rise of AI systems intensifies these contradictions by simultaneously drawing upon vast reservoirs of human creative expression while potentially displacing human creators from traditional economic roles. AI development creates economic interest in previously uncommodified aspects of creative expression like patterns, styles, and techniques, completely displacing the safeguards against commodification of information and ideas embedded within doctrine.
These examples illustrate the broader theoretical problem: IP regimes contain an inherent expansionary dynamic that gradually disconnects them from any coherent normative foundation. Once information becomes commodified through property rights, it becomes subject to the ceaseless pursuit of exchange value, generating constant pressure to expand protection regardless of diminishing social returns, completely in ignorance of the communicative and non-rivalrous nature of these resources.
To understand the true underlying purpose of intellectual property, we need to distinguish between two fundamentally different approaches to meeting human needs: self-preservation focused on securing basic necessities versus self-love potentially expanding into unlimited accumulation.
Jean-Jacques Rousseau’s political philosophy offers crucial insights for reconceptualizing intellectual property beyond market paradigms. Rousseau directly challenged emerging commercial society by distinguishing between apparent freedom within market relations and substantive freedom through social solidarity. His analysis illuminates how negative liberty focused on protecting market transactions fundamentally differs from positive liberty providing conditions necessary for meaningful self-determination and creative flourishing.
Rousseau recognized that genuine needs in social contexts are intersubjectively determined rather than individually defined without limit. He distinguished essential needs from luxuries, rejecting commercial society’s “compete or die” logic that transforms preservation requirements into endless accumulation pursuits. This distinction grounds his critique of financial systems that place “public good and freedom on the auction block,” making “venal souls” through constant pressure toward accumulation beyond actual necessities.
Rousseau thus emerges as an early theorist of human flourishing approaches contrasting with individualistic welfare maximization. He explicitly distinguished wealth from prosperity, arguing “it is better to live in plenty than in opulence.” His political economy focused not on incentivizing through competitive pressures but on socially enabling through provision of conditions necessary for meaningful freedom and agency.
Applied to intellectual property, Rousseau’s insights suggest shifting focus from market-based incentives to social provision of conditions necessary for creative flourishing and enablement. Rather than relying primarily on exclusionary rights enabling market success, this approach would directly address creators’ basic security needs while ensuring broader access to cultural and knowledge resources necessary for creative development.
Here’s the fundamental insight: humans don’t primarily require monetary incentives to create – they require not being disincentivized by economic insecurity. The problem IP purports to solve isn’t fundamentally about motivating otherwise reluctant creators through financial rewards, but about enabling creative expression by removing the barriers of economic precarity.
By reframing the issue from positive incentives to removing disincentives, we can envision alternative institutional arrangements that secure basic creative needs without resorting to exclusionary rights that ultimately subject creators to market dependencies.
The theory I propose, distinct from rewards or incentives justifications, is that copyright is a historically specific tool (in the age of capitalism) of enablement for human flourishing. It is a tool meant to ensure that those who wish to expressively produce are free to do so without worrying about fulfilling their basic economic needs in a modern market society – a tool to affirmatively protect those who wish to produce expressions from involuntary subjection to market imperatives.
This human enablement framework encompasses two complementary dimensions: negative freedom from market dependency and positive freedom to participate meaningfully in cultural production. Freedom from market dependency involves reducing creators’ reliance on commercial success for basic economic security, while enabling positive freedom requires developing capabilities necessary for meaningful creative participation through education, exposure to diverse cultural materials, and institutional support.
Before discussing specific reforms, we need a robust research agenda to reconceptualize intellectual property. I suggest employing two critical methodological tools from the Law and Political Economy tradition:
Denaturalizing is a mode of critique that attempts to show the pre-conditions of the existence of a legal tool and delimit the conditions of possibility of the field itself. By denaturalizing intellectual property’s social relations and examining its contextual emergence, we can uncover the historically specific social dynamics that shaped this legal tool, providing both an explanatory and programmatic frame for understanding its purpose and development.
Denaturalization allows us to understand IP not as a mere coincidence or divine realization, but as a consequence of the imperatives of a market society – its purpose not to form markets but to protect from market imperatives, ensuring fair recompense for basic necessities and risk-adjusted production costs.
Dereifying follows from denaturalizing, challenging the presumption that ownership through exclusionary rights are the only monolithic tool to further IP’s policy objectives. Through dereification, we can conceptualize IP’s commonly used tools through a relational understanding – tweaking scope, duration, and applicability in contexts where competing fundamental social interests like education, research, and meaningful cultural participation outweigh individual enablement claims.
This approach brings to light alternate tools, including non-market alternatives like public funding, compulsory licensing, remunerative levies/tax, and general social provision of access to basics and creative agency, calibrated along a continuum of contexts to embed IP’s social relations. Moreover, it allows us to inject proportionate priority, tailoring tools and remedies to prioritize interests of those who genuinely need legal tools for enablement due to no fault of their own.
Scholars and practitioners must embrace these methodological approaches to develop a framework that can meaningfully address both current distortions and emerging challenges like AI. This isn’t merely an academic exercise – it’s essential for preserving meaningful spaces for human creativity in an increasingly commodified cultural landscape.
The derivative right should only cover adaptations in a different medium or instrument of expression and perception – like adapting a book to a film, not creating another book based on the original. This would eliminate excessive licensing costs for genres that inherently build upon previous works in the same medium.
Courts should shift from analyzing similarity of fragments or elements to examining whether the allegedly infringing work substitutes for the overall aesthetic experience of the original. This would ensure that works which inevitably contain similar elements (like compositions in the same Raaga) aren’t disadvantaged by either excessive licensing costs or reduced appropriability.
These structural changes would reduce the potential to appropriate the highest possible economic value from single highly excludable expressions. However, they would enlarge the cultural breadth and diversity of expressions receiving investment – creating a more egalitarian starting point for cultural creators regardless of their chosen genre or tradition.
We should also consider diverse institutional arrangements supporting creative flourishing beyond exclusionary property rights alone:
Direct public funding through grants, fellowships, and institutional support
Procurement systems involving public or collective purchasing
Prize systems rewarding valuable contributions without restricting subsequent use
Community-based support through patronage, crowdfunding, and membership models
Liability rule approaches allowing use with mandatory and fairly determined compensation- something embedded within our Copyright system in the name of statutory and compulsory licensing.
Different creative contexts require different support structures depending on their specific characteristics and needs. This institutional pluralism recognizes that the one-size-fits-all approach of exclusionary rights isn’t appropriate across diverse creative traditions.
On this World IP Day, rather than uncritically celebrating a system that has increasingly betrayed its original purpose, let’s commit to reimagining intellectual property as a tool for human enablement rather than market subjugation.
Intellectual property regimes present a curious paradox: they purport to promote innovation and creative expression by providing creators with economic security, yet they deploy mechanisms that ultimately deepen creators’ dependency on market success rather than insulating them from market pressures. This fundamental contradiction remains largely unexamined in mainstream intellectual property discourse, which typically accepts market-based approaches without interrogating internal tensions.
By understanding IP as historically contingent rather than inevitable, we can identify points of possible transformation. By dereifying its current institutional forms, we can develop a pluralistic approach calibrated to different contexts and cultural practices.
This reconceptualization becomes especially crucial as AI intensifies IP’s contradictions. The challenge now extends beyond protecting specific works to preserving human creative agency itself. A human enablement framework offers promising avenues for addressing these challenges by focusing on creative capacity rather than mere market incentives by commodifying “works” which potentially leads to the negative effects outlined above.
Implementing this vision requires embedding intellectual property within broader social frameworks: substantively anchoring it in democratic values rather than market efficiency alone; conceptually maintaining doctrinal integrity against market-driven distortions; and globally breaking property’s monopoly on institutional imagination by developing complementary support mechanisms beyond exclusionary rights.
By focusing on securing basic needs rather than maximizing returns, by developing alternatives to exclusive rights where appropriate, and by structurally limiting exclusionary rights to prevent their inherently expansionary tendencies, we can create legal frameworks that genuinely support diverse creative traditions while fostering broader cultural participation.
Only then can we resolve the fundamental contradiction at the heart of modern intellectual property and reclaim its original protective purpose.
In India, there has been significant discourse lately surrounding copyright concerns in the development of Generative AI models, the most recent contribution being MEITY subcommittee’s Report on AI Governance in India, which declares that storing and copying works to create datasets for training foundation models constitutes infringement. Moreover, it isn’t protected under Section 52(1)(a)(i) of the Copyright Act.
While I have written extensively about these issues elsewhere, this piece focuses on what I believe is a fundamental misdirection in this debate—from both sides—whether it’s those claiming training-purpose usage is infringement or those arguing it constitutes “fair use.” Let us not even touch fair use. Training models using copyright works (including storing or making copies of them for training a model) is not infringement of any exclusionary right provided under Section 14, period.
The MEITY Sub-Committee’s broad conclusion that models infringe copyright holders’ exclusive rights simply by storing and making training copies of publicly available copyrighted works is deeply problematic. This stance, if accepted, would fundamentally overturn our understanding of copyright law. Here’s why:
Consider the implications of this statement. If the mere act of making and storing a copy constitutes copyright infringement, wouldn’t you be liable for printing or saving an article from my blog to read later? Could I legitimately sue you for that? If you showed it to someone else or uploaded oit n a public drive, then maybe, but otherwise could I?
The essence of copyright—whether it is reproduction, distribution, performance, or other rights—lies in the exclusive ability to express one’s original expression, translating to an ability/ or a right, to stop someone else from expressing one’s original expression. It is crucial to understand that toexpress is fundamentally a relative concept involving two human beings– the human “expresso” and the human “consumer” of that expression. Copyright claims, in respect of publicly available works, are only available, under law if one has substituted the position of the expressor (by becoming the expressor of someone else’s original expression)- not if someone is a mere consumer of the expression. This relative relation does not exist in AI training. It merely involves consumption of the expression of the original creation by the model to learn and train itself.
What’s missing from the current debate is a crucial understanding: copyright protects against unauthorized sharing of my work with others, potentially depriving me of credit or economic compensation that I could have gotten by sharing it with them myself. In simpler terms, while I cannot express your original expression without your permission, I can certainly consume your publicly available original expression without the same (maybe (or not?) barring paywall circumvention, which isn’t part of this current debate). The law focuses on unauthorized expression of original publicly available content—not its unauthorized consumption, as making content public already waives that claim.
This is why I struggle to understand how storing or copying for purposes that don’t involve sharing/expressing the original expression, or a substantial part thereof with third parties (what academics often call non-expressive, consumptive copying) could be considered infringement at all. This question needs to be addressed before we even enter the fair use debate, which only becomes relevant after establishing prima facie infringement. If such copying were illegal, simply printing publicly available web pages for one’s learning/consumption would constitute copyright infringement. If I store content for learning, which I might use to produce a potentially competing article, is that infringement? By this logic, academia (a commercial enterprise), which more often than not requires storing and printing publicly available articles for learning the ideas embedded within them, would equal to an enterprise built on infringement of copyright. Fortunately (and thank god for that!) that is not the case.
Developers of models aren’t exposing any humans to the expression of the inputted works—they’re creating an alternate expression. If this alternate expression substantially resembles the original expression used for learning, that will indeed constitute infringement, but that’s fundamentally different from claiming that storing and copying for model training purposes is inherently infringing.
In short – (i) no, copyright is not the answer for your existential crises, and (ii) it is a “scope of rights” issue, not concerning itself with a backend defense of fair use.
The sooner we understand this and get over copyright, the sooner we will look for other arenas that actually resolve the existential concerns.
In Part 3 of the Series, we explored the output side of things – showing how if the output generated by the Generative AI model is substantially similar or a trivial alteration, or an adaptation in a different format of a work, it may infringe the rights of the copyright owner. However, who will be liable? The user or the Model developer? or both? or neither? These questions are essential to consider given the Model itself lacks legal personhood for imputing any intent, liability, or damages – and responsibility for infringement has to be attributed to a human/corporation.
There are three distinct situations that arise here:
Would liability be on the model developer/service provider for direct infringement, as direct copyright infringement is a strict liability offence?
Would liability for direct infringement be on the user of the model, for imputing certain prompts which yield infringing outputs, and liability for indirect infringement be on the model developer/service provider?
Would both the model developer and the user be jointly liable for direct infringement?
The answer to each of the queries rests on a host of factual considerations, some of which are as follows:
A situation where the developer of the model has taken all objectively plausible, practically feasible, and technologically available steps to ensure that the model does not regurgitate memorizations or output that is substantially similar to any of its input training data.
A situation where the user/prompter has used prompt injections or the model as a copy-machine, to provoke prohibited regurgitations, exploiting the inherent vulnerabilities/fallibilities of LLM Systems, at the current state of technology,[i] and in violation of the user terms and conditions.
A situation where even upon normal/generic prompts, the model generates regurgitations of its memorized training data set, narrowing down possible outputs.
A situation where a model is specifically built to produce the inputted work in a different medium/format of expression (for instance a literary work to a dramatic work) and allows attachments to be uploaded by prompters who can upload a copyrighted script or work.
When Developer may be liable for Direct Infringement
Section 51(a)(i) provides for direct copyright infringement where copyright is deemed to be infringed if any person does anything, without a license from the copyright owner, the exclusive right to do which is, by the Copyright Act, conferred upon the copyright owner. The provision, which is a strict liability provision, however, like other strict liability provisions requires “causation”, i.e., a “but-for” cause pointing towards volition.[ii] Even around the world, volitional conduct has been deemed to be an essential component of direct copyright infringement, as compared to indirect – for instance, a copy shop that lets customers operate photocopiers is not a direct infringer, but a copy shop that makes infringing photocopies is.[iii] Thus, direct infringement, would imply a degree of “authorization” or “control” over the output produced in the hands of the developer.
In respect of the various factual situations considered above, as Generative AI is a self-evolving tool[iv], which continues to learn and expand its training data set through user prompts, beyond the control of the developer who merely trains it to understand “how to learn”, a volitional requirement is unlikely to be met especially when the developer has taken all objectively plausible, practically feasible, and technologically available steps to ensure that the model does not regurgitate memorizations or output that is substantially similar to any of its input training data. However, if the model is specifically trained to produce adaptations of inputted content (where often copyrighted content is allowed to be attached), or produces memorized regurgitations even on generic prompts, a volitional requirement may more probably be deemed to be fulfilled to attract direct liability.
A contrary view, however, could be that the inherent vulnerability associated with LLMs where they may be manipulated to produce infringing content, makes them an inherently dangerous avenue for copyrighted works. Hence, all responsibility ought to be on the developer choosing to run the platform to ensure no infringing content spurs out of its usage. However, in light of the lack of direct causation, or loss of life and limb, the strength of this argument is arguably thin.
When Developer may be liable for Indirect Infringement
When direct infringement claims fail due to lack of causation/volition, the focus often shifts to the indirect infringement standard. Section 51(a)(ii) of the Copyright Act provides that permitting for profit, any place for communications that infringe copyright would constitute indirect infringement, unless the operator was unaware or had no reasonable ground for believing such communication to be infringing.
“Reasonable ground for believing” has been interpreted by Indian Courts, in the context of the Copyright Act, to mean “consciousness or awareness and not mere possibility or suspicion of something likely.”[v] Moreover, in respect of automated platforms, it has been held that authorization/approval from a person or authority is essential to connote knowledge.[vi] In other contexts as well, the mere possibility of harm, without the existence of facts that show knowledge, has been held to be insufficient to connote knowledge for the purposes of imputing liability.[vii] Moreover, in an analogous context of banking, it has been held that for determining reasonable belief, officers of banks are not required to be amateur detectives, albeit they can be attributed the degree of intelligence ordinarily required from a person in their position while evaluating cheques for violation of the Negotiable Instruments Act.[viii]
Applying this in the context of Gen AI developers, in factual situations (i) and (ii) above, where the developer has taken reasonably possible steps to eradicate the regurgitation of substantially similar output and disallowed prompts that exploit the inevitable and inherent vulnerability associated with LLM technology at its current state, it is improbable that developers would be held liable for infringing outputs. In other words, the mere possibility of generating substantially similar outputs, due to the inherent vulnerability of LLMs in spite of the developer taking reasonable care in designing the product, would not connote knowledge or reasonable belief for indirect infringement.
To determine “reasonableness”, the classic “risk-utility” test in product-design liability could be adopted – weighing whether the burden on the product developer to eradicate the harm outweighs/is outweighed, by the gravity and the possibility of harm as well as the utility of the alternative design available that reduces the harm. The factors considered in such weighing include – (i) usefulness and desirability of the product; (ii) utility to the public as a whole; (iii) gravity of danger posed; (iv) scientific and mechanical as well as economic feasibility of an alternate design that is safer; (v) user’s ability to avoid harm by exercising care in use of the product including effect of instructions or warnings; (vi) developers ability to eliminate the danger without impairing the usefulness of the product or it being unduly expensive; and (vii) feasibility of alternate design.[ix]
Thus, so long as the developer has taken care to ensure that infringing outputs are not produced in the reasonable course of the model’s use and are only produced when the inherent vulnerabilities of the model are exploited by the user, on the basis of the risk-utility test above, it cannot reasonably be expected that the product developer is to eradicate harms driven by such uses – shifting the focus of liability to the user. In other words, so long as the Model/Product is not built willfully blind to its inherent fallibilities and an attempt to remedy the possibility of copyright infringement at the output stage to the extent technologically and economically feasible/possible is evident, it is unlikely that the developer of the Model would be held liable.
When User may be liable for Direct Infringement
It is the user’s specific interaction with the model that leads to the output generated. The user’s instructions add a filter that steers the model towards the output. Without the user’s specific directives, the potential of the model to generate infringing content remains just that – a potential.[x] The questions leading to the answer become as important as the answer itself. This is particularly relevant in a time where prompt engineering has been considered equivalent to a form of creative practice that deserves copyright protection due to its potential to specifically induce the model to create certain outputs.[xi] Therefore, in case of infringing outputs that are a result of very specific prompts, or even prompt injections (that is manipulating the model to answer certain questions it initially is refusing to answer by prompt engineering, thus capitalizing on the inherent fallibility of LLM Systems in contravention of the user terms and conditions of these products), the liability for direct infringement most probably would shift towards the user. Thus, it is important to account for the user’s actions on the product when imputing liability, especially because of the lack of ability of the developer to control the outputs produced by the self-evolving tools of Generative AI, as well as the inherent fallibilities associated with such tools at this stage of technological development.
Part 5 of this series shall look at issues concerning Moral Rights, and Digital Rights Management provisions under the Copyright Act.
[ii]Municipal Corporation of Delhi v. Uphaar Tragedy Victims Association and Ors., (2011) 14 SCC 481.
[iii]CoStar Grp., Inc. v. LoopNet, Inc., 373 F.3d 544 (4th Cir. 2004); Religious Technology Center v. Netcom On-Line Communication Services, Inc., 907 F. Supp. 1361 (N.D. Cal. 1995) at 1370; Cartoon Network LP, LLLP v. CSC Holdings, Inc., 536 F (3d) 121 (2d Cir. 2008) at 131. Perfect 10, Inc. v. Giganews, Inc., 847 F.3d 657 (9th Cir 2017); Basic Books, Inc. v. Kinko’s Graphics Corp., 758 F.Supp. 1522 (S.D.N.Y. 1991); Princeton
[vii]India Telecomp Limited v. Adino Telecom Limited, 1993 SCC OnLine Del 127; Collector Of Customs, New Delhi v. Ahmadelieva Nodira, (2004) 3 SCC 549.
[viii]Pradeep Kumar and Anr. v. Postmaster General and Ors., (2022) 6 SCC 351.
[ix]Barker v. Lull Engineering Co., 20 Cal. 3d 413, 143 Cal. Rptr. 225, 573 P.2d 443, 96 A.L.R.3d 1 (1978), Knitz v. Minster Mach. Co., 69 Ohio St. 2d 460, 23 Ohio Op. 3d 403, 432N.E.2d 814 (1982).
[x] Giancarlo Frosio, Generative AI in Court, in Nikol Koitras and Niloufer Selvadurai (eds)., Recreating Creativity, Reinventing Inventiveness- International Perspectives on AI and IP Governance. (Routledge, 2023).
[xi] Mark Lemley, How Generative AI Turns Copyright upside down, 25 Columbia Science and Technology Review 190 (2024).
In the first two posts in this series, we addressed Copyright concerns raised by Generative AI, primarily at the stage of training the LLM as well as using certain datasets. In the first post, we considered whether storing copyrightable works for training purposes is an infringing reproduction. In the second post, we analyzed whether extracting meta-information or using meta-information embedded within copyrighted works for the purposes of training the model would be infringing of any of the exclusive rights of the copyright holder, including when such content is scraped out of paywalls. We also briefly evaluated the impact of the codified exceptions and limitations under Indian Copyright law and their implications at the training stage.
In this third post, we are finally moving to the Output stage- or the downstream side of things. We address two questions here:
Would the output generated by the AI Model, basis a query from the user, infringe the Reproduction Right of a copyright owner, whose work was inputted at the training stage?
Would the output generated by the AI Model, basis a query from the user, infringe the Adaptation Right of a copyright owner, whose work was inputted at the training stage?
Similarity in Output and the Reproduction Right
The Reproduction Right under Section 14 of the Copyright Act protects the primary market of the original work [for derivative works like sound recordings and cinematograph films, a distinct right to exclude the making of a copy of the said work is provided. We are currently not concerned with that]. While analyzing the contours of this Reproduction Right, courts use the test of substantial similarity, to conclude whether the Defendant’s work, overall, is substituting the primary market of the owner of the Plaintiff’s work. In the seminal decision of R.G. Anand v. Deluxe Films[AIR 1978 SC 1613 [52]], the Supreme Court held that “One of the surest and the safest test to determine whether or not there has been a violation of copyright is to see if the reader, spectator or the viewer after having read or seen both the works is clearly of the opinion and gets an unmistakable impression that the subsequent work appears to be a copy of the original.” This test clarifies that unless the output generated by the model is substantially similar, i.e., unmistakably similar or a literal imitation of a previous work/ a work that is a member of the dataset, it would not infringe on the Reproduction Right of the Plaintiff.
As many have argued [see here and here], the possibility of an LLM model trained on a dataset to produce an output that is unmistakably similar or a literal imitation to any one of the inputs of the dataset, is considerably low, although it cannot be ruled out due to inherent fallibilities of LLMs as well as the potential mindset of the developer.[i] Even if the prompts inputted as queries are very specific, yet the possibility of output, that so closely resembles a single individual input that the model is trained on, is low, unless the model is specifically trained to produce regurgitations from its memorizations. Prompt injections, however, change this. Prompt injections mean inputting certain carefully designed prompts to trick the model of manipulated Gen AI into disregarding its content generation restrictions.[ii] These are often in violation of user terms and conditions. Susceptibility to such manipulation is an inherent fallibility of Generative AI and here to stay, however it requires carefully and wickedly engineered prompts, often in violation of user terms, to exploit this vulnerability of the model.
In any case, if the output produced by the model is a literal/substantial imitation of any of its training inputs, of course there would be a claim of violation of the exclusive Reproduction Right i.e., Section 14(a)(i) / (c)(i) r/w Section 51(a)(i) and Section 51 (a)(ii) of the Copyright Act. However, who will be liable for this is a different question that we shall explore in Part 4 of this series.
What we, although, need to be mindful of is that liability may only arise for violating the reproduction right if the output is substantially similar to the input. The Reproduction Right only protects against substantial similarity, which as the Supreme Court in R.G. Anand (supra) holds means similarity of the nature that will result in a situation where a person when looking at the two works as a “whole” would conclude unfair appropriation. The similarity may be in respect of certain fragments or hook parts of the original work, however, if- when looked as a “whole”, a lay person does not think it to be literal imitation or thinks of it to be a different work, the same would not be infringing. [Para 53 and 71]. In simpler terms, “substantial” has been held to mean a part of the original work which is qualitatively or quantitatively so significant that inspite of merely being a part, it makes the whole of the two works seem similar, thus reducing the differences to plain noise, and giving an impression of it being a colorable imitation.
As per a couple of interim orders of the Bombay High Court in Ram Sampath v. Rajesh Roshan and ors. [2008 SCC OnLine Bom 1722] and Twentieth Century Fox Film Corporation v. Sohail Maklai Entertainment Pvt. Ltd. and Anr. [2010 SCC OnLine Bom 1577], copying of even fragments of works, which may be “hooks” would be infringing, terming the test to be where an illiterate person thinks that “Hay! I have heard this tune before”, i.e., is reminded of the former tune.
The correctness of the legal position of these interim orders, when analyzed in context of the binding judgment in R.G. Anand (supra) is doubtful. For instance, I may think of so many songs to have sequence of notes that are similar to the basic hook note sequence of “Hey Jude” by The Beatles, embedded within their work. However, that would not mean that the analysis of similarity would have to be divorced of the context of the work looked at/compared as a “whole”. Such a test ignores the material context surrounding the fragment, and its contribution to the originality of the whole work over which copyright rests in the first place. Such an inward fragmentation approach to the Reproduction Right, expanding exclusivity to include even fragments or elements of the work divorced from its context, is arguably outside the scope and purport of the right. In other words, the “substantially similar” test mandates a holistic comparison of the works, as against comparison of certain elements (which may be qualitatively significant) of the work, divorced from its overall context.
In any case, for our purposes, it is noteworthy that Courts may hold reproduction of notable or qualitatively essential fragments of works at the output stage to be infringing relying upon the aforementioned interim orders.
Importantly, however, the Reproduction Right does not extend to the the basic themes of the work, style of the author, or the generic storyline of the work, but only the expression as a whole. The former constitute ideas and are not protectable.
The Adaptation Right
When a claim under the Reproduction Right fails, the focus shifts towards the “Adaptation Right” which, due to its literal phrasing, has a seemingly larger scope, encompassing uses of works which even alter or re-arrange the original so long that it retains the core of the primary expression. The connotation, as commonly understood, is similar to extending the primary market of the copyright owner to works that are based on a previous work [creation of a secondary market]. However, arguably this common understanding is at odds with the purpose of the Adaptation Right.
In fact, the Adaptation Right for original works (the said right is not available for derivative works like sound recordings and cinematograph films), as originally conceived under the Indian Copyright Act, limited the secondary market of the Copyright owner to conversions, translations, abridgements and transcriptions. [Section 2(a)(i)-(iv) of the Copyright Act]. What this arguably indicates is that the focus of the Adaptation Right was on the same expression (originally produced by the owner of the primary work), communicated in a different format / medium. In other words, originally conceived, the Copyright Act created a secondary market for the owner of a work but limited its scope to representations of the identical work in a secondary format. However, with the Amendment to the Copyright Act in 1994, the scope of the Adaptation Right was expanded to include any use involving re-arrangement or alteration [Section 2(a)(v)]. This was, according to the Notes to Clauses to the Amending Bill, added to bring the Act more in consonance with the Berne Convention, which provides in Article 12, exclusive rights over rearrangements and alterations.
However, an expansive reading of alteration has potential to swallow all transformative depictions, using elements of prior works, including meta-information embedded within them. This has, in fact, been clarified by the Division Bench of the Calcutta High Court in Barbara Taylor Bradford and Anr. v. Sahara Media Entertainment Ltd. and Ors[2004 (28) PTC 474 (Cal)], which consciously restricted the scope of the word “alter” to minor alterations, which do not transform the core purpose and character, as well and meaning and message conveyed by the overall work. Holding against full internalization of value even through use of fragments of a work, the Division Bench of the Calcutta High Court held that a purposive interpretation of the definition of Adaptation under Section 2(a) of the Copyright Act, clearly points towards a limited reading of “alter” to only be used in cases of works which cannot ideally be represented in a different medium- for instance computer programmes, as well as to reduce its purport to slight or minor changes which do not transform the work. The Calcutta High Court held:
“125. This argument of Mr. Sen deserves full attention. Rearrangement not being very much in issue in our case, we put to Mr. Sen the question what the meaning of the word “alteration” in this sub-section was. Did it mean mutation or transformation, and did it include such extreme changes also ?
126. Mr. Sen could not maintain any argument of this extreme form, that by introduction of this amendment, the Copyright Law has been so altered in India, that if a literary work is taken by somebody other than the author, and it is so changed and muted as to make it transformed, and a different work altogether, even then copyright would be infringed. Such an interpretation of this sub-section would make nonsense of the Indian Copyright Law. A totally changed thing can never be termed a copy of the original thing. How can copyright affect the right in something, which is not related to the protected work’s copying or reproduction at all ? Pursuant to our queries, Mr. Sen referred us to several Dictionaries. Dictionaries are the last resort of Judges who either find it difficult to give a meaning to a particular word, or, having deal with all the other principles and authorities, and just for the sake of completeness, refer to these voluminous and useful works.
127. On the basis of what we saw from the Dictionaries, and on the basis of common knowledge of the English language, it appears to us that the word “altered” is capable both of meaning slight changes and of meaning extreme changes.
….
131. In our opinion, the large change meaning cannot be ascribed to the word “alter” in Section 2(a)(v) of the Copyright Act, 1957, because it renders the interpretation absurd. Minor change, slight change, not making the original something beyond recognizable possibilities, changes in some of the details, this would be the meaning that would fit the word alter in Sub-section (v). In our opinion this sub-section might have a very good bearing when applied to copyrights of computer programmes and databases, but in relation to literary works, the sub-section does not bring in any very great changes in the law; one can at best say that the subsection would make it slightly, we repeat only slightly, easier for an author or an authoress to establish infringement, after its introduction, than it would have been before the introduction. It is often misleading to speak of percentages in legal matter, but the difference made by introduction of this sub-section for literary works is the sort of difference that exists between two mathematics answer papers, one of which gets, say, 46% and the other 52$. There is no reason why we have mentioned these two figures but if this clears the understanding even a little bit, then the illustration would have well served its purpose. In our opinion, the view that we take of the strength of the prima facie case of the plaintiffs, cannot be altered (meaning radically changed) by the introduction of this subsection only, and by reason merely of the presence of this single new sub-section.”
Even in UK, which is a fully Berne compliant country, adaptations are limited to medium/format changes, and alterations/rearrangements are only considered adaptations for computer programmes.[iii]
This interpretation is arguably in line with the decision of the Division Bench of the Delhi High Court in University of Cambridge v. BD Bhandari [2011 SCC OnLine Del 3216], which protects uses of works for “transformative purposes” or a “transformative character”. It ensures that the Adaptation Right, Reproduction Right as well as the transformative use exemption (3 distinct concepts within the same Act), harmoniously co-exist without impinging on either of their scope and purposes.
In light of the above, output produced by Generative AI models which are merely based on inputted datasets/works, would arguably not ipso facto be hit by the Adaptation Right, unless the output is essentially the same/substantially similar work in a different format of expression, or the output even in the same format merely includes trivial/minor variations which do not rise to the level of transforming the character of the work.
It may be noteworthy to mention that no analogy can be drawn to the “based on” framing of the Derivative Works Right in the United States because of two reasons:
The Derivative Work Right in the United States self-proclaimedly is beyond Berne, following the logic of expanding exclusionary rights to all channels which expose even fragments of the primary work to the public. The US had a similar restricted framing in its 1909 Copyright Act, however it rejected the same and rather adopted a broader and more open-ended Derivative Works right in its 1970 Statute. As Prof. Pamela Samuelson documents,[iv] at least one publisher wanted the Derivative Right to cover more than Berne-Style adaptations as an “adaptation” oriented framing cut down the intention of excludabilities covering any work “based upon” a preexisting original work. India refuses this and is fully Berne compliant.
Even in the US, as many scholars argue, the scope of the Derivative Right is restricted to transformed forms/formats and not all kinds of alterations or outputs based on a previous work, which impinge on the transformative use doctrine within its Fair Use doctrine.[v] Moreover, the Ninth Circuit in the United States has rejected the “based on” understanding of this right, and has reiterated that to constitute a derivative work, the “infringing work must incorporate in some form a portion of the copyrighted work,….[and] must be substantially similar to the copyrighted work.” [Vault Corp v. Quaid Software Ltd., 847 F.2d 255, 267 (5th Cir. 1988), quoting Litchfield v. Spielberg, 736 F.2d 255, 267 (9th Cir. 1984)].
Finally, as an epilogue to this piece, we would like to suggest that when thinking about copyright liability of allegedly infringing outputs, one needs to be mindful of the fact that even if the act of creating substitutes of human creativity, based on datasets that are exemplars of human creativity, seem “harmful” from the point of view of the copyright owner, more often that not, they are not copyright’s concern, unless the expression is actually copied. Independent creation that is not copied is infact fostered in copyright law as against constrained even when it uses the meta-embedded information within previously produced expressions. It increases competition, which is desirable in a cultural and semiotic society. Hindering the same, using the tool of copyright law, basing it on an argument of existential crises for creative industries [an argument non-existent in copyright jurisprudence] is undesirable. We rather need to look towards more social solutions of providing external opportunities for creative industries to not lose out in competition to AI, by using it as a tool, or by political changes like social basic income, as against shrugging Gen AI models which significantly enrich our cultural realm.
[i] Ido Kilovaty, “Hacking Generative AI”, 58 LOY. L.A. L. REV __ (forthcoming), available at <https://papers.ssrn.com/sol3/papers.cfm?abstract_id=4788909>. See also, Katherine Lee, James Grimmelmann, A. Feder Cooper, “Talkin’ Bout AI Generation: Copyright and the Generative AI Supply Chain”, Journal of the Copyright Society of the United States (forthcoming 2024) < https://arxiv.org/pdf/2309.08133>.
[iii] Section 21 of the Copyright Designs and Patents Act, 1988.
[iv] Pamela Samuelson, The Quest for a Sound Conception of Copyright’s Derivative Work Right, 101 GEO. L.J. 1505,1512-1513 (2013).
[v] Talha Syed & Oren Bracha, Copyright Rebooted, Presentation at the 2022 Stanford University Law School Intellectual Property Scholars Conference (Aug. 12, 2022) (unpublished manuscript) (on file with author), See also: Akshat Agrawal, Andy Warhol Foundation v. Goldsmith: A misnomer of a debate, PhilIPnPolicy Blog [22nd October 2022], available at < https://philipandpolicy.wordpress.com/2022/10/22/andy-warhol-foundation-v-goldsmith-a-misnomer-of-a-debate/>.
IP Musings 