Artificial intelligence entered the newsroom quietly. The first automated news reports - short, formulaic summaries of financial data and sports scores - attracted little attention precisely because they were indistinguishable from routine human output. Today, the scale is different. The Associated Press alone uses automated systems to generate thousands of financial and earnings reports every quarter,1 and platforms across India and the world deploy natural language generation tools to produce content at speeds and volumes no human team could match.2
The legal system has not kept pace. When a journalist writes an article, the Copyright Act, 1957 assigns authorship and ownership through a framework developed for human creative labour. When a machine writes the same article, that framework breaks down entirely. There is no author in the statutory sense. There is no work-for-hire relationship of the kind the Act contemplates. There is, in short, a legal vacuum - and it is one that courts and legislators in India have yet to seriously confront.3
This vacuum is not merely academic. It has immediate, practical consequences. If AI-generated journalistic content is not protected by copyright, news organisations have no legal recourse against rivals who copy and republish their machine-generated output wholesale. If it is protected, the question of who holds that protection - the developer of the AI, the journalist who prompted it, or the organisation that deployed it - remains entirely unresolved under current Indian law. Either outcome carries significant risks for the economics of journalism and the public interest in a functioning news ecosystem.
This article proceeds as follows. Part II situates AI-generated journalism within its technological context and draws a critical distinction between AI-assisted and AI-generated content. Part III examines where and why the Copyright Act, 1957 breaks down when applied to machine-generated works. Part IV analyses the three principal competing claims to authorship and ownership, and explains why each fails under existing law. Part V draws comparative lessons from the United Kingdom and the United States. Part VI advances the central argument of this article: that India needs a sui generis protection framework tailored to AI-generated journalistic content. Part VII addresses the two strongest counterarguments, and Part VIII concludes with a specific legislative recommendation.
The scope of this article is deliberately narrow. It addresses AI-generated journalistic content specifically - written news articles, reports and summaries produced by automated systems with minimal human creative input - and focuses on the Indian copyright framework, with comparative reference to the UK and US positions.
Before engaging with the law, a foundational distinction must be drawn. Not all journalism that involves AI is AI-generated journalism. When a reporter uses an AI tool to transcribe interviews, suggest headlines, or identify patterns in a dataset, the final article remains the product of human intellectual effort. The AI functions as an instrument - no different, in principle, from a word processor or a search engine. The copyright analysis in such cases is straightforward: the human author retains authorship.
The position changes fundamentally when the AI system itself generates the substantive content of the article. Natural language generation platforms - fed structured data such as corporate earnings figures, match scores, or election results - produce complete, publication-ready articles with no meaningful human creative contribution at the output stage. A journalist may have designed the template, selected the data source, or triggered the process, but the text itself is the machine’s. That is the category of content this article addresses.
Examples are no longer hypothetical. Bloomberg’s Cyborg system generates thousands of news stories from financial data each year.4 Automated Insights’ Wordsmith platform has been widely used by American media organisations, and similar tools are increasingly deployed in Indian digital newsrooms for cricket scores, stock market updates, and election result breakdowns. The volume of such content is growing. Its legal status in India remains, at present, entirely unclear.
That ambiguity matters for three reasons. First, news organisations that invest significantly in AI infrastructure have a legitimate interest in knowing whether their output is legally protected. Second, without clear ownership rules, the market for AI-generated journalism is susceptible to free-riding - competitors copying machine-generated content without licence or payment. Third, and most broadly, the question of who owns machine-generated news has direct implications for press diversity, media economics, and ultimately the public’s access to information.
The Copyright Act, 1957 is a statute designed for a world in which only humans create. Every major provision governing authorship, ownership, and protection rests on the implicit assumption that there is a person - a natural or legal person - at the origin of any protected work. That assumption, reasonable in 1957, is no longer universally accurate.
Section 2(d) of the Act defines ‘author’ in relation to literary works as ‘the author of the work.’5 The circularity of this definition has been noted in academic commentary, but the courts have consistently interpreted it to require a human being exercising intellectual skill and judgment. The Supreme Court of India in Eastern Book Company v D B Modak6 held that copyright subsists only in works that are the product of the author’s own intellectual effort - works reflecting ‘minimum degree of creativity’ rather than ‘mere labour or industry.’7 While the Court did not address AI, the logic of its reasoning is clear: creativity, in the statutory sense, presupposes a human mind giving expression to original thought.
Section 2(o) defines ‘literary work’ to include ‘computer programmes, tables and compilations including computer databases.’8 Some commentators have argued that this provision is broad enough to extend copyright protection to AI-generated text - reasoning that such content is, in some sense, a product of computer processing. This argument does not withstand scrutiny. Section 2(o) defines what kinds of works qualify as literary works; it says nothing about who qualifies as their author. A computer-generated article may be a literary work in form, but if there is no qualifying author, section 13 - which conditions copyright on authorship9 - denies it protection entirely.
Section 17 addresses ownership and vests first copyright in the author, subject to exceptions for works made in the course of employment.10 The employer-employee exception requires an identifiable human author whose work was made ‘in the course of the author’s employment under a contract of service.’11 No such author exists in the case of fully AI-generated content. The machine is not an employee. Its developer is not the author of every output the machine produces. And the news organisation that deployed it has no statutory basis for ownership in the absence of an underlying author.
The combined effect of these provisions is a structural gap. AI-generated journalistic content, if produced without meaningful human creative input at the output stage, does not fit comfortably within any existing category of protected work under the Copyright Act, 1957. It risks being classified as unprotected - part of the public domain by default, free for anyone to copy and republish without restriction. That is not a trivial outcome. It is, for the media industry, a potentially existential one.
When AI generates a news article, three parties might plausibly claim authorship or ownership: the developer who built the AI system, the user or journalist who prompted it, and the news organisation that deployed it. Each claim has some intuitive appeal. Each fails under existing Indian law.
The Developer’s Claim. The most technically grounded claim is the developer’s. The AI system is the developer’s creation, and its outputs, one might argue, are a downstream expression of the developer’s intellectual architecture.12 There is a superficial logic to this. But it does not survive close examination. Copyright protects specific expression, not the capacity to produce expression. A developer who builds a platform capable of generating an unlimited variety of articles has not authored any of those articles in the way that copyright law requires. The developer exerts no creative control over the specific words produced in response to a given data input. To hold otherwise would give AI developers a perpetual and unlimited copyright interest in every output their systems ever generate, an outcome that would be both practically unworkable and conceptually unjustifiable.
The User’s or Journalist’s Claim. The second claim - that the journalist or editor who prompted the AI system is the author of its output - rests on the idea that creative choices were made in selecting the data, framing the query, or refining the result.13 This argument is strongest where there is substantial human creative input in the prompting process. It is weakest - and most relevant to this article - where the AI system operates largely autonomously on structured data. When a system ingests a corporate earnings release and outputs a 400-word report without any meaningful human creative direction at that stage, the prompter’s contribution is too thin to sustain an authorship claim under the Modak standard. Selecting a dataset is not the same as authoring its textual expression.
The News Organisation’s Claim. The most commercially significant claim is that of the deploying news organisation, which has invested in the AI infrastructure and bears the financial risk of its operation. This claim has strong policy support - news organisations need the incentive of legal protection to justify investment in AI journalism - but it has no adequate statutory basis under current Indian law. Section 17’s employer exception requires an underlying human author in an employment relationship. Without one, the news organisation cannot rely on it. There is simply no provision in the Copyright Act, 1957 that vests ownership in a legal entity on the basis of deploying a creative machine.
The failure of all three claims has a single, important consequence: AI-generated journalistic content, as the law currently stands in India, may be entirely unprotected. It enters the public domain the moment it is published - not by legislative design, but by default, because the statute never contemplated its existence. This is not a defensible outcome. It penalises investment, rewards free-riding, and provides no incentive for news organisations to develop and deploy AI responsibly.
Two jurisdictions offer the most instructive comparative material: the United Kingdom, which has the only statutory provision in the common law world explicitly addressing computer-generated works, and the United States, where the Copyright Office and federal courts have recently confronted AI authorship directly. Neither jurisdiction has fully resolved the problem, but both illuminate the path India must take, and the dead ends it should avoid.
The United Kingdom. Section 9(3) of the Copyright, Designs and Patents Act 1988 provides that in the case of a computer-generated work, ‘the author shall be taken to be the person by whom the arrangements necessary for the creation of the work are undertaken.’14 This is a remarkable provision, and a lonely one. No other major common law jurisdiction has enacted anything comparable. It resolves the authorship question by legislative fiat, vesting authorship in the person who made the arrangements for the work’s creation, regardless of whether that person made any direct creative contribution.
The UK provision is not without its difficulties. The Court of Appeal in Nova Productions Ltd v Mazooma Games Ltd15 declined to resolve precisely what ‘arrangements’ means in this context, and the question of whether section 9(3) can apply to outputs generated by sophisticated machine learning models, as opposed to simpler algorithmic systems, has not been authoritatively settled. More recently, in Thaler v Comptroller-General of Patents,16 the UK Supreme Court declined to extend analogous reasoning in the patent context to allow a machine to be named as an inventor, signalling judicial conservatism even in the face of technological change. Section 9(3) nonetheless remains valuable as proof of concept: a legislature can, through purposive drafting, resolve the AI authorship problem without waiting for the courts to do so.
The United States. The United States Copyright Office has taken an increasingly firm position. In its 2023 guidance on AI and copyright,17 the Office confirmed that it will not register copyright in works produced entirely by AI systems without human authorship. The federal courts have reached similar conclusions. In Thaler v Vidal,18 the Federal Circuit held that ‘inventors must be natural persons’, and while the case concerned patent law, its reasoning has been applied by analogy to copyright claims. The Supreme Court’s decision in Feist Publications Inc v Rural Telephone Service Co19 had already established that originality, the sine qua non of copyright, requires human authorship.
The US position is clear, but it is also static. It tells us what the current law does not protect; it does not tell us what the law should do. Congress has not acted, and the Copyright Office has acknowledged that legislative intervention may ultimately be necessary.20
The lesson from both jurisdictions is the same: existing copyright frameworks were not designed for machine authorship, and courts are poorly placed to remedy that deficiency through interpretation alone. Legislative action is required. The UK’s section 9(3) demonstrates that such action is possible. The US experience demonstrates what happens when it is deferred.
The argument of this article is not that AI-generated journalistic content should receive the same copyright protection as human-authored works. It is that such content needs a different kind of protection entirely, one designed for its specific characteristics rather than force-fitted into a framework built for human creativity. The appropriate model is a sui generis regime: a purpose-built set of rights, limited in scope and duration, that addresses the distinct legal and economic realities of machine-generated news.
Why Amendment Alone Is Insufficient. The most obvious legislative response would be to amend section 2(d) of the Copyright Act, 1957 to include AI systems within the definition of ‘author.’ This approach, while superficially straightforward, creates more problems than it solves. Granting full copyright to AI-generated works, with a term of sixty years from publication under the current Act21, would confer an extraordinarily long monopoly on content produced at negligible marginal cost and without the human creative labour that copyright has historically been designed to reward and incentivise. It would also create perverse incentives: news organisations would have a strong financial interest in minimising human editorial involvement in order to secure the broadest possible IP protection for the longest possible term.
A second amendment strategy, extending section 2(o)’s definition of computer-generated works, faces the same objection. It would plug a definitional gap but leave the fundamental question of appropriate protection scope and duration unaddressed. Piecemeal amendment of a framework that was not designed for this problem is an inadequate response to it.
The Structure of a Sui Generis Regime. This article proposes a sui generis framework modelled, in its structural logic, on the European Union’s database protection directive22, which recognised that databases, though not meeting the originality threshold for conventional copyright, nonetheless represent significant investment deserving of legal protection. The analogy is instructive: AI-generated journalistic content, like a database, reflects substantial investment in infrastructure and data without necessarily reflecting creative authorship in the traditional sense.
The proposed framework has four components. First, ownership should vest automatically in the news organisation that deploys the AI system to generate the content. This is the entity that has made the relevant investment, bears the financial risk, and is best placed to enforce rights and answer for misuse. The developer’s contribution, building the system, is already protected by separate intellectual property rights in the software itself. The prompter’s contribution, where minimal, does not justify independent ownership. Organisational ownership is the most commercially coherent and practically administrable solution.23
Second, the term of protection should be substantially shorter than that afforded to human-authored works. A ten-year term from the date of first publication is proposed. This is sufficient to allow news organisations to recover their investment and maintain a competitive advantage in the market for their AI-generated output. It is not so long as to remove machine-generated content from the public domain for a period disproportionate to the investment it represents. The EU database directive’s fifteen-year term24 provides a useful reference point, though the specific characteristics of journalistic content, which depreciates rapidly in news value, justifies a shorter period.
Third, any AI-generated journalistic content must carry a mandatory disclosure. Where an article or report has been produced by an automated system without meaningful human authorship, that fact should be clearly stated, in a standardised form, prescribed by regulation, at the point of publication. This requirement serves two distinct purposes. It preserves reader trust by ensuring that audiences know when they are reading machine-generated content. And it creates a clear evidentiary record that demarcates AI-generated from human-authored works, reducing the scope for uncertainty in enforcement proceedings.25
Fourth, upon expiry of the protection term, AI-generated journalistic content should pass automatically into the public domain. There is no justification, economic or normative, for perpetual or extended protection of machine-generated news. The content has served its commercial purpose. Its entry into the public domain serves the broader public interest in the free flow of information and the historical record of public affairs.
Institutional Implementation. A sui generis framework of this kind would require amendment to the Copyright Act, 1957 through the insertion of a new chapter specifically addressing AI-generated works. The Copyright Office of India would need to develop a registration mechanism, optional but evidentially valuable, for AI-generated content, alongside prescribed disclosure standards. The Law Commission of India, which has previously recommended legislative modernisation in response to technological change,26 is the appropriate body to develop the detailed legislative proposal. Parliamentary intervention, informed by that recommendation, is the necessary final step.
This framework deliberately stops short of addressing AI-generated content in fields other than journalism. The specific characteristics of news content, high volume, low creative input, rapid depreciation in value, strong public interest in access, make it a distinct category warranting distinct treatment. Extending the analysis to AI-generated literary fiction, advertising copy, or academic writing raises different considerations beyond the scope of this article.
‘The Existing Law Is Flexible Enough.’ The first objection to this article’s argument is that Indian courts, applying a purposive interpretation to the existing Copyright Act, 1957, could extend protection to AI-generated works without legislative intervention. On this view, the judiciary is capable of adapting the statute to new technological realities, as it has done in other areas of intellectual property law.
This objection underestimates the specificity of the problem. The Modak standard of minimum creativity is not merely a threshold, it is a framework premised on the existence of a human author whose intellectual choices are being assessed. It cannot be applied to a machine output without fundamentally distorting its meaning. No amount of judicial creativity can produce from section 2(d) a provision that vests authorship in a legal entity on the basis of deploying an AI system. The statutory text does not permit it. Courts are rightly reluctant to legislate from the bench on questions of this magnitude, and reliance on judicial interpretation in the absence of legislative action would produce inconsistent, unpredictable outcomes across different High Courts, precisely the kind of legal uncertainty that the media industry can least afford.27
‘AI Content Needs No Legal Protection.’ The second objection holds that AI-generated journalistic content, precisely because it is produced at low cost and high speed, does not require the incentive of copyright protection. The argument is that, unlike human creative labour, machine output will be produced regardless of whether it is legally protected, and that copyright would therefore confer a windfall on news organisations without any corresponding public benefit.
This argument misunderstands how the incentive structure of media investment works. News organisations invest in AI journalism infrastructure, in data licensing, in system development, in editorial oversight, because they expect to be able to exploit the output commercially. That exploitation depends, in part, on having enforceable legal rights against free-riding competitors. Without such rights, the rational response is underinvestment in AI journalism infrastructure, which ultimately reduces the volume and diversity of automated news coverage. The public interest case for protecting AI-generated journalistic content is therefore not merely about rewarding investment, it is about maintaining the conditions under which that investment continues to be made.
India’s Copyright Act, 1957 cannot answer the questions that AI-generated journalism now poses. Its authorship framework presupposes human creativity. Its ownership provisions require a human author at their foundation. Its originality standard, as interpreted by the Supreme Court in Eastern Book Company v D B Modak, cannot be applied to machine output without conceptual distortion. The result is a structural legal vacuum, one in which AI-generated journalistic content is left authorless and unprotected, not by design, but by default.
Three competing ownership claims, by the AI developer, the user-journalist, and the deploying news organisation, each fail under current law, for reasons this article has examined in detail. Comparative analysis of the UK and US positions confirms that even the most legally developed jurisdictions are struggling with the same problem, and that judicial interpretation alone cannot resolve it. Legislative intervention is not merely desirable, it is necessary.
This article has argued for a sui generis protection framework: one that vests ownership automatically in the deploying news organisation, limits protection to ten years from first publication, mandates clear disclosure of AI origin, and defaults to the public domain upon expiry. This framework draws structural inspiration from the EU’s database protection directive, adapts it to the specific characteristics of journalistic content, and is designed to balance the legitimate interests of news organisations in recovering their investment against the broader public interest in the free flow of information.
The Law Commission of India should be tasked with developing the detailed legislative proposal. Parliament should act on that recommendation before the courts are forced to fill the vacuum through improvised interpretation - with all the inconsistency and unpredictability that such improvisation entails. The technology will not wait. The law should not either.
1. Associated Press, “AP Continues to Invest in Automated Technology” (AP, 2023) <> accessed 15 May 2026.
2. Francesco Marconi, Journalism's New Calling: The Rise of AI in the Newsroom (Columbia University Press 2020) 12–14.
3. Lumen Database, “Automated Journalism and Copyright Disputes” (2023) <https://lumendatabase.org> accessed 11 May 2026.
4. Copyright Act 1957, s 2(d).
5. Eastern Book Company v D B Modak (2008) 1 SCC 1 (Supreme Court of India).
6. ibid.
7. Copyright Act 1957, s 2(o).
8. Copyright Act 1957, s 13(1)(a).
9. Copyright Act 1957, s 17.
10. Rupendra Kashyap v Jiwan Publishing House AIR 1996 Del 258 (Delhi High Court).
11. Ryan Abbott, “I Think Therefore I Invent: Creative Computers and the Future of Patent Law” (2016) 57 Boston College Law Review 1079, 1092.
12. Annemarie Bridy, “Coding Creativity: Copyright and the Artificially Intelligent Author” (2012) Stanford Technology Law Review 5, 9–11.
13. Thaler v Vidal 43 F 4th 1207 (Fed Cir 2022)
14. United States Copyright Office, “Copyright and Artificial Intelligence” (Part 1: Digital Replicas, 2023) <https://copyright.gov/ai> accessed 1 March 2025.
15. Copyright, Designs and patents Act 1988 (UK), s 9(3)
16. Nova Productions Ltd v. Mazooma Games Ltd. [2007] EWCA Civ 219 [24] (Court of Appeal).
17. Thaler v. Comptroller-General of Patents, Designs and trade Marks [2023} UKSC 49 [52]-[55].
18. Directive 96/9/EC of the European Parliament and of the Council of 11 March 1996 on the Legal Protection of databases [1996] OJ L77/20.
19. Regulation (EU) 2024/1689 of the European parliament and of the Council of 13 june 2024 laying Down Harmonised Rules on Artificial Intelligence (AI Act) [2024] OJ L1689/1, art 4.
20. Jane Ginsburg and Luke Budiardjo, “Authors and Machines” (2019) 34 Berkeley Technology law Journal 343, 381-383.
21. Andres Guadamuz , “Do Androids Dream of Electric Copyright?” (2017) 2 Intellectual Property Quarterly 169, 184.
22. Robert Yu, “The Machine Author: What Level of Copyright Protection Is Appropriate for Fully AI-Generated Text?” (2016) 164 University of Pennsylvania Law Review 2245, 2265.
23. Copyright Act 1957, s 22.
24. Directive 96/9/EC (n 18) art 7.
25. Law Commission of India, “Speak for Yourself” (Report No 185, 2003).
26. Digital Personal data Protection Act 2023.
27. Feist Publications Inc v Rural Telephone Service Co 499 US 340 (1991) 345.