The global legal architecture governing intellectual property is currently facing its most severe conceptual challenge since the Industrial Revolution. The source of this upheaval is the emergence of advanced, multi-modal Generative Artificial Intelligence (AI) frameworks capable of autonomously producing high-fidelity literary texts, musical scores, complex software code, fine artistic imagery, and complex industrial designs. Historically, the legal justification for granting monopolistic intellectual property rights has been rooted in natural law and utilitarian theories, both of which presume a human creator. For instance, John Locke’s labor theory suggests that an individual acquires property rights by mixing their physical or mental labor with raw resources.1 Similarly, utilitarian theory positions IPR as an economic incentive system designed to encourage human beings to innovate for the broader welfare of society. Autonomous algorithmic systems challenge these foundational doctrines, as they produce highly innovative outputs without human emotional investment or labor.
In the Indian context, the statutory legal regime was built upon this strictly human-centric baseline. When the Copyright Act, 1957, and the Patents Act, 1970, were drafted, the concept of a machine acting as an independent creator or inventor belonged entirely to the domain of speculative science fiction.2 Consequently, the language embedded within these statutes repeatedly enforces the necessity of a human agent. Today, however, AI platforms do not merely act as passive instruments like a paintbrush or a typewriter; instead, they make independent decisions regarding execution, composition, and selection, often minimizing the human role to a simple natural language prompt. This shift creates a problematic legal vacuum in India: if an autonomous machine creates a work that would otherwise easily meet the legal threshold for patentability or copyright protection, who owns that work? Denying protection altogether could stifle commercial investment in AI development, while granting it to the human user or the AI engine could dilute the statutory rights meant to safeguard human creators.
The core obstacle to recognizing AI-generated works under the Indian Copyright Act, 1957, lies within the explicit definition of an ‘author’ under Section 2(d). The statute enumerates specific definitions based on the nature of the work. In relation to a literary or dramatic work, the author is ‘the author of the work’; in relation to a musical work, it is ‘the composer’; in relation to an artistic work other than a photograph, it is ‘the artist’; and in relation to a photograph, it is ‘the person taking the photograph’. Since the 1994 amendment, Section 2(d)(vi) has also provided that, in relation to a literary, dramatic, musical or artistic work which is computer-generated, the author is ‘the person who causes the work to be created’.3 Even this clause vests authorship in a person behind the machine rather than in the machine itself. The persistent use of the word ‘person’ in these clauses has traditionally been read as referring to a natural person or, in specific commercial contexts, a legally recognized corporate entity. This human prerequisite is further reinforced by Section 2(qq), which defines a ‘performer’ as including an actor, singer, musician, dancer or ‘any other person who makes a performance’, a performance being a visual or acoustic presentation made live by one or more performers.4 These statutory boundaries make it exceptionally difficult to designate an AI algorithm as an author under existing Indian law.
This statutory barrier is further complicated by judicial standards regarding originality. For a long period, Indian courts leaned towards the UK’s common-law doctrine of ‘sweat of the brow’, which granted copyright protection based on the mere expenditure of labor, skill, and capital.5 However, in the landmark case of Eastern Book Company v. D.B. Modak (2008), the Supreme Court of India shifted the legal benchmark significantly.6 The Court rejected the simplistic ‘sweat of the brow’ approach and instead adopted a middle-path standard closely aligned with the Canadian ‘skill and judgment’ test.7 The Supreme Court held that for a work to be protected as original, it must demonstrate a minimal degree of creativity and must not be a product of mere mechanistic labor.8 The Court held:
“To claim copyright in a compilation, the author must produce the material with exercise of his skill and judgment which may not be creativity in the sense that it is novel or non-obvious, but at the same time it is not a product of merely labour and capital. The derivative work produced by the author must have some distinguishable features and flavour to raw text of the judgments delivered by the court. The trivial variation or inputs put in the judgment would not satisfy the test of copyright of an author.”9
This standard creates an interesting paradox when applied to Generative AI. The process by which an AI model processes trillions of parameters to construct a new image or write an essay can be described as entirely mechanistic and algorithmic. While the final output may appear highly creative to a human reader, the underlying process lacks the conscious, intellectual choice required by the D.B. Modak standard. Because the machine operates via statistical probability and pattern replication, its output is fundamentally a ‘purely mechanical exercise’,10 thereby failing the legal test for originality. Consequently, fully autonomous AI works are unlikely to attract copyright in India: Section 2(d) recognizes no machine as an author, and even output claimed under the computer-generated works clause must meet the judicial insistence on a conscious ‘human touch’.
Similar structural hurdles exist within the domain of patent law. Under the Indian Patents Act, 1970, the right to file a patent application is governed by Section 6, under which an application for a patent for an invention may be made by ‘any person claiming to be the true and first inventor of the invention’, by that person’s assignee, or by the legal representative of a deceased person so entitled.11 While the term ‘true and first inventor’ is defined in Section 2(1)(y), the definition only states who it excludes, failing to provide an affirmative definition. However, Section 2(1)(p) explicitly defines a ‘patentee’ as the ‘person’ for the time being entered on the register as the grantee or proprietor of the patent,12 and the entire text of the Act consistently connects inventorship with human attributes, such as mental conception, technical intent, and legal accountability.
This legal stance has been tested globally through the high-profile Artificial Inventor Project, under which Dr. Stephen Thaler filed patent applications worldwide naming his autonomous AI system, DABUS (Device for the Autonomous Bootstrapping of Unified Sentience), as the sole inventor. The DABUS applications were consistently rejected by the United States Patent and Trademark Office (USPTO), the European Patent Office (EPO), and the UK Intellectual Property Office (UKIPO).13 The definitive legal position in the United Kingdom was articulated by the UK Supreme Court in Thaler v. Comptroller-General of Patents, Designs and Trade Marks (2023), where the court ruled that under the UK Patents Act 1977, an inventor must be a natural person.14 The court held that DABUS, being a machine, is not a person at all, and that Dr. Thaler’s ownership of the machine gave him no right to apply for or obtain patents for what it produced; it noted that the outcome might well have been different had he claimed to be the inventor himself, using DABUS as a highly sophisticated tool.15
No Indian court has yet decided the question, but the Indian Patent Office has now followed the same analytical path: on 15 April 2026 it refused Dr. Thaler’s Indian DABUS application, holding that only a natural person can be recognized as an inventor under the Patents Act, 1970, and Dr. Thaler’s appeal against that refusal is pending before the Delhi High Court.16 Under the current framework of the Patents Act, 1970, an invention requires an ‘inventive step’ under Section 2(1)(ja), which demands a technical advance or economic significance that makes the invention ‘not obvious to a person skilled in the art’.17 The standard of non-obviousness is measured against the cognitive capacity of a human researcher. Because an AI system does not possess cognitive intent and cannot hold legal title or execute assignments, it cannot be legally recognized as the ‘true and first inventor’ under Section 6. As a result, autonomous AI inventions fall into the public domain by default in India, creating a major commercial risk for research firms utilizing AI for pharmaceutical drug discovery or advanced materials engineering.
Beyond the question of ownership, Generative AI models present an immediate risk regarding copyright infringement during their training phase. Modern Large Language Models (LLMs) and diffusion models require the ingestion of massive datasets containing millions of copyrighted books, academic articles, photographs, and artistic works. This process of data scraping involves making temporary or permanent digital copies of protected works into the AI’s training memory without obtaining prior licensing agreements from the copyright owners. Under Section 14 of the Indian Copyright Act, 1957, the copyright owner possesses the exclusive right to reproduce the work in any material form, including storing it in any medium by electronic means.18 Therefore, the unauthorized copying of protected works to train AI models constitutes a prima facie violation of the copyright owner’s exclusive reproduction rights.
AI developers frequently attempt to defend this practice by invoking the doctrine of ‘fair dealing’ codified under Section 52(1)(a) of the Indian Copyright Act.19 This provision exempts certain acts from infringement if they are carried out for the purposes of private or personal use (including research), criticism or review, or the reporting of current events and current affairs. However, the Indian concept of fair dealing is significantly narrower than the open-ended ‘fair use’ doctrine utilized in the United States under 17 U.S.C. § 107.20 In Wiley Eastern Ltd. v. Indian Institute of Management (1995), the Delhi High Court described the basic purpose of Section 52 as the protection of freedom of expression under Article 19(1) of the Constitution, so that research, private study, criticism, review and the reporting of current events could be protected.21 Indian law has no single, uniformly applied test of fairness: whether a dealing is fair is a question of fact and degree, and in its recent interim ruling on AI training the Delhi High Court framed the inquiry around whether the use is confined to its stated purpose, whether it competes with the work or damages the owner’s legitimate interests, and the public interest.22
Because most leading AI models are developed by commercial entities to create competitive commercial products, their reliance on Section 52(1)(a) remains contested. At the interim stage in ANI Media Pvt. Ltd. v. Open AI OpCo LLC, the Delhi High Court held in July 2026 that commercial use does not by itself exclude the defense, and that OpenAI’s storage of a news agency’s articles to train its large language models was, prima facie, a fair dealing for the purpose of research; that view is provisional, and an appeal against it is pending.23 Nonetheless, if an AI model is trained on a contemporary Indian artist’s style and subsequently generates images that directly compete with that artist’s commercial market, the use is unlikely to be considered ‘fair’. Furthermore, unlike the US, where courts in cases like Authors Guild v. Google, Inc. (2015) ruled that mass digital copying for search and data analysis can be fair use if it is highly transformative,24 Indian law lacks a general statutory exception for Text and Data Mining (TDM).25 Consequently, commercial AI training practices in India remain exposed to substantial infringement claims from authors, publishers, and collective management organizations while the question remains unsettled.
To formulate an effective legislative solution for India, it is essential to analyze how other major jurisdictions are responding to the intersection of AI and IPR. The global legal landscape is currently divided into three distinct regulatory models, as summarized below:
| Jurisdiction | Copyright Authorship | Patent Inventorship | Data Mining Exceptions |
|---|---|---|---|
| United States | Strictly human. The Copyright Office refuses registration for machine-authored or fully AI-created works.26 | Strictly human. Affirmed in Thaler v. Vidal; AI cannot be named as an inventor.27 | Highly flexible. Heavily relies on the judicial doctrine of ‘Fair Use’ (17 U.S.C. § 107) for transformative commercial models.28 |
| United Kingdom | Hybrid model. Sec. 9(3) CDPA protects computer-generated works, granting 50-year protection to the human arranger.29 | Strictly human. Affirmed by the Supreme Court in Thaler v. Comptroller-General.30 | Very limited. Strict exception for non-commercial research; no general commercial text and data mining exception.31 |
| European Union | Strictly human. Requires the work to reflect the ‘author’s own intellectual creation’ (ECJ standard).32 | Strictly human. Rejected by European Patent Office rules requiring a human inventor.33 | Statutory protection. Article 3 of the DSM Directive permits TDM for scientific research by research organizations and cultural heritage institutions; Article 4 permits TDM for any purpose, including commercial use, unless rightholders have expressly reserved their works.34 |
Table 1: Comparative regulatory approaches to AI and intellectual property
Given the limitations of the current frameworks under the Copyright Act, 1957, and the Patents Act, 1970, India requires an immediate legislative intervention. Applying standard copyright protections to AI outputs risks diluting human creativity, while a complete denial of protection leaves tech investments highly vulnerable. Therefore, India should introduce a specialized, sui generis (unique) legal framework tailored specifically for AI-generated works.
First, the legislature should build on the existing computer-generated works clause in Section 2(d)(vi) of the Copyright Act by creating a separate category of ‘Computer-Generated Works’, modeled loosely on Section 9(3) of the UK’s CDPA, but with strict variations.35 The ownership of an autonomous AI output should be granted to the person ‘by whom the arrangements necessary for the creation of the work are undertaken’. However, the duration of this right must be significantly shorter than standard human copyright, reducing it from the traditional 60 years after the author’s death36 to a brief term of 10 to 15 years from the date of creation. This ensures that tech companies can recoup their financial investments without permanently locking up public domain expressions.
Second, to address the data scraping challenge, India should implement an explicit ‘Opt-Out’ Text and Data Mining statutory exception. Under this provision, commercial AI developers would be legally permitted to scrape data for training purposes, provided they pay a standardized statutory licensing fee to a centralized collective management organization (akin to IRRO or IPRS).37 Authors and publishers would retain the absolute right to formally ‘opt out’ of AI training datasets via digital metadata tagging. If a developer uses an opted-out work, they would face severe statutory damages. This dual framework protects human creative rights while providing a transparent, predictable path for AI innovation in India.
The intersection of Artificial Intelligence and Intellectual Property Rights represents a profound conceptual turning point for Indian jurisprudence. As demonstrated throughout this study, the human-centric focus of existing statutes like the Copyright Act, 1957, and the Patents Act, 1970, is fundamentally incompatible with the realities of modern generative models. The judicial standards established in landmark decisions such as Eastern Book Company v. D.B. Modak affirm that copyright protection requires a distinct ‘human touch’ and conscious intellectual effort, elements that mathematical algorithms cannot replicate.38 Similarly, patent frameworks refuse to grant inventor status to non-human entities lacking legal personhood.
To maintain its position as an emerging global tech hub while fiercely safeguarding its rich community of human creators, India cannot rely on slow, incremental judicial interpretations. The legislative recommendations detailed in this paper, ranging from a shortened sui generis copyright term for computer-generated works to a robust statutory data licensing scheme, offer a balanced path forward. By modernizing its legal framework, India can foster technical innovation, guarantee fair compensation to human authors, and create a highly structured, stable marketplace for the future of digital content.
*****
1. See John Locke, Two Treatises of Government bk. II, ch. V, § 27 (1690).
2. The Copyright Act, No. 14 of 1957, India Code (1957); The Patents Act, No. 39 of 1970, India Code (1970).
3. Copyright Act, supra note 2, § 2(d)(i)–(iv), (vi). Sub-clauses (v) and (vi) were substituted in their present form by The Copyright (Amendment) Act, No. 38 of 1994, India Code (1994), § 2, with effect from May 10, 1995.
4. Copyright Act, supra note 2, § 2(q) (“performance”, in relation to performer’s right, means “any visual or acoustic presentation made live by one or more performers”), § 2(qq).
5. Burlington Home Shopping Pvt. Ltd. v. Rajnish Chibber, 61 (1995) DLT 6 (Del.) (India) (protecting a compilation of customer details as a literary work on the strength of the skill and labor invested in it).
6. Eastern Book Co. v. D.B. Modak, (2008) 1 SCC 1 (India) (decided Dec. 12, 2007).
7. CCH Canadian Ltd. v. Law Soc’y of Upper Can., 2004 SCC 13, [2004] 1 S.C.R. 339, ¶ 16 (Can.); Eastern Book Co., supra note 6, ¶¶ 37–38 (applying the Canadian standard under the Copyright Act, 1957).
8. Eastern Book Co., supra note 6, ¶¶ 38, 40.
9. Id. ¶ 38.
10. Id. ¶ 37 (summarizing CCH Canadian, supra note 7, ¶ 16: the exercise of skill and judgment “must not be so trivial that it could be characterized as a purely mechanical exercise”).
11. Patents Act, supra note 2, § 6(1).
12. Id. § 2(1)(p), (y).
13. Thaler v. Vidal, 43 F.4th 1207 (Fed. Cir. 2022) (affirming the USPTO’s refusal; an inventor under the Patent Act must be a natural person), cert. denied, 143 S. Ct. 1783 (2023); Case J 8/20, Designation of Inventor/DABUS (EPO Legal Bd. of Appeal Dec. 21, 2021) (an inventor designated under Article 81 and Rule 19(1) EPC must be a person with legal capacity, not a machine); Decision BL O/741/19 (UK Intell. Prop. Off. Dec. 4, 2019).
14. Thaler v. Comptroller-General of Patents, Designs & Trade Marks [2023] UKSC 49, ¶ 56 (appeal taken from Eng.) (decided Dec. 20, 2023).
15. Id. ¶¶ 52, 56, 90.
16. Indian Patent Application No. 202017019068 (Stephen L. Thaler), refused by the Indian Patent Office on Apr. 15, 2026 (holding that only natural persons can be recognized as inventors under the Patents Act, 1970, and also finding no inventive step); Ambika Aggarwal, The Inventor Is Still Human: Indian Patent Office’s DABUS Refusal, SpicyIP (Apr. 21, 2026), https://spicyip.com/2026/04/the-inventor-is-still-human-indian-patent-offices-dabus-refusal.html; S.N. Thyagarajan, Can AI Be Considered Inventor under Patent Law? Delhi HC Seeks Patent Office Response to US Scientist’s Plea, Bar & Bench (July 30, 2026), https://www.barandbench.com/news/litigation/can-ai-be-considered-inventor-under-patent-law-delhi-hc-seeks-patent-office-response-to-us-scientists-plea (notice issued on Dr. Thaler’s appeal against the refusal).
17. Patents Act, supra note 2, § 2(1)(ja).
18. Copyright Act, supra note 2, § 14(a)(i) (the right “to reproduce the work in any material form including the storing of it in any medium by electronic means”).
19. Id. § 52(1)(a). The Explanation to clause (a) provides that “[t]he storing of any work in any electronic medium for the purposes mentioned in this clause . . . shall not constitute infringement of copyright.”
20. 17 U.S.C. § 107.
21. Wiley Eastern Ltd. v. Indian Inst. of Mgmt., 61 (1996) DLT 281, ¶ 19 (Del.) (India).
22. ANI Media Pvt. Ltd. v. Open AI OpCo LLC, I.A. 45300/2024 in CS(COMM) 1028/2024, ¶¶ 234–236 (Del. HC July 24, 2026) (India), https://www.scobserver.in/wp-content/uploads/2026/08/DHC-OpenAI-v-ANI-Judgement.pdf (no single test has been adopted uniformly or consistently by Indian courts; counsel broadly agreed that the US four-factor test is not applicable in India).
23. Id. ¶¶ 176–179, 190, 219, 256, 271–274 (findings expressly prima facie, with “no bearing on the final outcome of the suit”); Riya Rathore, Delhi High Court Issues Notice on ANI Appeal Against Rejection of Interim Injunction Plea Against OpenAI, LiveLaw (Sept. 15, 2026), https://www.livelawbiz.com/copyright/delhi-high-court-issues-notice-on-ani-appeal-against-dismissal-of-interim-injunction-plea-in-copyright-suit-against-openai-550080.
24. Authors Guild v. Google, Inc., 804 F.3d 202 (2d Cir. 2015).
25. Dep’t for Promotion of Indus. & Internal Trade, Working Paper on Generative AI and Copyright, Part 1: One Nation One License One Payment: Balancing AI Innovation and Copyright 14 (Dec. 2025), https://www.dpiit.gov.in/static/uploads/2025/12/ff266bbeed10c48e3479c941484f3525.pdf (“There is currently no specific exception under copyright law for text and data mining . . . .”).
26. Thaler v. Perlmutter, 130 F.4th 1039 (D.C. Cir. 2025), cert. denied, No. 25-449 (U.S. Mar. 2, 2026); U.S. Copyright Office, Copyright and Artificial Intelligence, Part 2: Copyrightability iii (Jan. 2025), https://www.copyright.gov/ai/Copyright-and-Artificial-Intelligence-Part-2-Copyrightability-Report.pdf (copyright “does not extend to purely AI-generated material, or material where there is insufficient human control over the expressive elements”).
27. Thaler v. Vidal, supra note 13.
28. 17 U.S.C. § 107; Authors Guild, supra note 24.
29. Copyright, Designs and Patents Act 1988, c. 48, §§ 9(3), 12(7) (UK); Dep’t for Sci., Innovation & Tech. & Dep’t for Culture, Media & Sport, Report on Copyright and Artificial Intelligence 14 ¶ 47, 111 ¶ 53 (Mar. 2026), https://assets.publishing.service.gov.uk/media/69ba692226909a14239612e4/CP2602959_-_Report_on_Copyright_and_Artificial_Intelligence_web.pdf (proposing, “in the absence of evidence of its ongoing value”, that the protection for wholly computer-generated works be removed).
30. Thaler v. Comptroller-General, supra note 14, ¶ 56.
31. Copyright, Designs and Patents Act 1988, supra note 29, § 29A (copies for text and data analysis for non-commercial research); Report on Copyright and Artificial Intelligence, supra note 29, at 11 ¶ 27 (a broad copyright exception with opt-out “is no longer the government’s preferred way forward”).
32. Case C-5/08, Infopaq Int’l A/S v. Danske Dagblades Forening, ECLI:EU:C:2009:465, ¶ 37 (July 16, 2009).
33. Case J 8/20, supra note 13. The EPO applies the European Patent Convention and is not an institution of the European Union.
34. Directive (EU) 2019/790 of the European Parliament and of the Council of 17 April 2019 on Copyright and Related Rights in the Digital Single Market and Amending Directives 96/9/EC and 2001/29/EC, arts. 3–4, 2019 O.J. (L 130) 92.
35. Copyright Act, supra note 2, § 2(d)(vi); Copyright, Designs and Patents Act 1988, supra note 29, § 9(3).
36. Copyright Act, supra note 2, § 22.
37. Compare DPIIT Working Paper, supra note 25, exec. summary (recommending, by majority, a mandatory blanket licence for AI training with a statutory remuneration right, collected by a central non-profit body of copyright societies and collective management organisations, under which rightholders would not have the option to withhold their works), with Report on Copyright and Artificial Intelligence, supra note 29, at 11 ¶ 27.
38. Eastern Book Co., supra note 6, ¶¶ 38, 40.