VidhiAagaz
International Journal ofLaw Management & Humanities
ISSN 2581-5369 · Open access
International Journal ofLaw Management & Humanities
Open accessArticlesVol. 9 No. VI (2026)pp. 1146–1165

Who Benefits from Machine Invention? The Right to Science, Artificial Intelligence and the Indian Patent System

0views0downloads
Published 19 November 2026 · pp. 1146–1165 · Vol. 9 No. VI (2026)
Abstract

Debates on artificial intelligence (AI) and patent law have concentrated on two questions: whether an AI system can be an inventor, and whether AI-related inventions fall within patentable subject matter. A third question has been largely neglected: who benefits from AI inventions, and on what terms? This article examines that question through the human right to participate in and enjoy the benefits of scientific progress and its applications, recognised in Article 27 of the Universal Declaration of Human Rights and Article 15 of the International Covenant on Economic, Social and Cultural Rights, and authoritatively interpreted by the Committee on Economic, Social and Cultural Rights in General Comment No. 25 (2020). The article argues that AI patents have a dual character. They can advance the right to science by attracting investment, compelling disclosure and diffusing useful applications in health, disability access and education. They can also undermine it by concentrating control over AI in a small number of firms, restricting access to essential applications, limiting follow-on research and offering only thin disclosure of opaque models. Patent filings in generative AI have grown sharply and are dominated by a handful of large enterprises, which sharpens these concerns. The article then shows that Indian law already contains the materials to read the right to science into patent practice. These include Articles 21, 51(c) and 51A(h) of the Constitution; the purposive approach of the Supreme Court in Novartis; the principles in Section 83 of the Patents Act, 1970; the exclusions in Sections 3(b) and 3(k); the disclosure requirement in Section 10 as elaborated by the Computer Related Inventions Guidelines, 2025; the research exemption in Section 47; and the compulsory licensing and government use provisions. The article concludes with recommendations for examination practice, disclosure standards, research exemptions, access mechanisms and a human rights impact lens in AI patent policy.

Keywords
Right to ScienceArtificial IntelligencePatentsGeneral Comment No. 25Human RightsPatents Act, 1970Access to TechnologyCompulsory LicensingDisclosure
Contents
Full text

Introduction

Artificial intelligence is now one of the most heavily patented fields of technology. The World Intellectual Property Organization (WIPO) documented about 54,000 generative AI patent families filed between 2014 and 2023, with more than a quarter of them published in the final year alone.1 Its 2026 update reports that more generative AI patent families were published in 2024 and 2025 combined than in the entire preceding decade.2 India is among the top five inventor locations for generative AI patents.3

The legal response to this growth has followed two main tracks. The first concerns inventorship. Courts and patent offices across the world have considered whether an AI system can be named as an inventor, and have almost uniformly answered that it cannot. The Indian Patent Office reached the same conclusion in April 2026 when it refused the application naming the AI system DABUS as inventor.4 The United States Patent and Trademark Office has confirmed that AI is a tool and that the ordinary conception standard governs all inventions.5 The second track concerns patentable subject matter, where India’s Section 3(k) exclusion for computer programmes “per se” and algorithms has been progressively clarified by courts and by the Computer Related Inventions (CRI) Guidelines, 2025.6 Abroad, the UK Supreme Court’s decision in Emotional Perception has realigned UK law with the approach of the European Patent Office.7

Both tracks ask whether AI inventions can be protected and by whom. Neither asks the question that matters most to the public: who benefits from AI inventions, and on what terms? Patents are not ends in themselves. They are instruments that society adopts to encourage innovation and the disclosure of knowledge. Whether the patent system serves that purpose for AI depends on whether its outputs reach the people who need them, and whether it supports or obstructs further research.

International human rights law offers a framework for asking that question. Article 27 of the Universal Declaration of Human Rights (UDHR) and Article 15 of the International Covenant on Economic, Social and Cultural Rights (ICESCR) recognise the right of everyone to share in scientific advancement and its benefits.8 In General Comment No. 25 (2020), the Committee on Economic, Social and Cultural Rights (CESCR) gave this “right to science” its first authoritative interpretation.9 The General Comment addresses intellectual property directly, and expressly identifies artificial intelligence as an emerging technology that can both enhance rights and intensify inequality.10

This article applies that framework to AI patents, with particular reference to India. It addresses three questions:

1.  What does the right to science require of States in relation to intellectual property and emerging technologies such as AI?

2.  In what ways do AI patents advance, and in what ways do they threaten, the enjoyment of that right?

3.  How can the right to science be given effect within the existing framework of Indian patent law?

The article proceeds as follows. Part II sets out the content of the right to science and the obligations it imposes. Part III describes the AI patent landscape and the Indian legal framework. Parts IV and V examine, respectively, how AI patents advance and threaten the right. Part VI identifies the constitutional and statutory tools through which Indian law can give effect to the right. Part VII offers recommendations, and Part VIII concludes.

The article makes two contributions. First, it moves the AI-patent debate beyond inventorship and eligibility to the distribution of benefits, which has received little attention in Indian scholarship. Second, it shows that India does not need a new statute to begin this work; the Patents Act, 1970 and the Constitution already provide the tools, if they are read in light of the right to science.

The right to science: content and obligations

A. Textual foundations

Article 27(1) UDHR recognises the right of everyone to share in scientific advancement and its benefits, while Article 27(2) protects the moral and material interests of authors of scientific productions.11 Article 15(1) ICESCR translates these into binding obligations: the right to take part in cultural life, the right to enjoy the benefits of scientific progress and its applications, and the right to benefit from the protection of the moral and material interests of authors. Article 15(2) to (4) add duties to conserve, develop and diffuse science, to respect the freedom indispensable for scientific research, and to encourage international scientific cooperation.12 India acceded to the ICESCR on 10 April 1979.13

For decades this right remained, in the CESCR’s own words, one of the areas of the Covenant to which States give least attention.14 General Comment No. 25 sought to change that.

B. Benefits and participation

General Comment No. 25 gives “benefits” a wide meaning. Benefits include the material results of applied research, the scientific knowledge and information that research produces, and the role of science in forming critical, informed citizens.15 Importantly, the right is not limited to passive receipt. The Committee rejects a rigid division between scientists who produce science and a public entitled only to consume it. The right is to participate in as well as to enjoy the benefits of scientific progress.16

The Committee identifies four interrelated elements, alongside freedom of research. Table 1 sets them out with their relevance to AI patents.17

Element Meaning under General Comment No. 25 Relevance to AI patents
Availability Scientific progress takes place and its applications are protected and widely disseminated; States should promote open science Whether patents promote or retard dissemination of AI knowledge
Accessibility Equal access to applications, especially where needed for other rights; access to information on risks; open opportunity to participate Price and licensing of essential AI applications; access to data and models for researchers
Quality Most advanced, verifiable science; regulation and certification for responsible use Reproducibility and adequacy of patent disclosure for AI models
Acceptability Science explained and applied in culturally appropriate ways; ethical standards; special protection of vulnerable groups Inventions that profile or discriminate; needs of persons with disabilities and children

Table 1: Elements of the right to science under General Comment No. 25 and their relevance to AI patents

C. The right to science is not a right to patents

A common misconception is that Article 15(1)(c) ICESCR establishes a human right to intellectual property. The CESCR rejected this reading in General Comment No. 17 (2005), distinguishing the human right of authors, which protects the personal link between creators and their creations, from most legal entitlements recognised in intellectual property systems.18 General Comment No. 25 reaffirms that distinction.19 The UN Special Rapporteur in the field of cultural rights put the point directly in her 2015 report on patent policy: the right to science and culture does not establish a human right to patent protection, but it does provide a human rights framework within which patent policy must be considered.20 The same report states that where patent rights and human rights conflict, human rights must prevail.21

This has a direct implication for AI. Patents are policy instruments that States may calibrate. Their legitimacy depends on whether they serve the ends that the right to science protects.

D. Intellectual property in General Comment No. 25

General Comment No. 25 treats intellectual property as having a complex, two-sided relationship with the right.22 On one side, patents provide economic incentives that draw private actors into research. On the other, the Committee identifies three ways in which intellectual property can harm the right:23

1.  Distortion of research funding, as private investment flows only to profitable projects and neglects areas vital for economic, social and cultural rights.

2.  Restrictions on the sharing of information, such as data exclusivity, which hinder the advancement of science.

3.  Barriers to access, because patents allow holders to set prices that may put essential applications beyond the reach of low-income persons and developing countries.

In response, the Committee asks States to foster the positive effects of intellectual property while avoiding its negative effects. It recommends public funding and alternative incentives, such as rewards that delink remuneration from sales, and it asks States to guarantee the “social dimensions” of intellectual property in national law.24 The Committee describes intellectual property as a social product with a social function.25 In the context of health, it states that States should use, when necessary, all the flexibilities of the TRIPS Agreement, including compulsory licences.26

E. AI and emerging technologies

General Comment No. 25 expressly names artificial intelligence among emerging technologies that are blurring the boundaries between the physical, digital and biological worlds.27 The Committee notes that AI can produce large gains in productivity and efficiency, but that algorithms can also reinforce discrimination and that these technologies may intensify social inequality.28 It calls for three things: enhanced international cooperation; decision-making on these technologies within a human rights framework, including mechanisms so that autonomous systems avoid discrimination, can explain their decisions and allow accountability; and regulation of the ownership and control of data according to human rights principles, given that many emerging inequalities are linked to the capacity of some businesses to exploit massive data.29

F. State obligations

The right imposes obligations to respect, protect and fulfil.30 The obligation to protect requires States to prevent private actors from interfering with the right, for example by preventing access to knowledge.31 Among the core obligations, States must eliminate laws, policies and practices that unjustifiably limit access to goods and information related to science, and ensure access to applications of science critical to the right to health and other rights.32 At the level of national implementation, the Committee asks States to adopt a normative framework that includes measures to harmonise intellectual property with the right to science.33 It also confirms that the right is justiciable.34

These obligations form the benchmark against which the AI patent system is assessed in the rest of this article.

AI innovation and the patent system

A. The global landscape: growth and concentration

Two features of the AI patent landscape matter for the right to science: rapid growth and concentration of ownership.

On growth, WIPO’s 2024 landscape report recorded that generative AI patent families rose from 733 in 2014 to more than 14,000 in 2023.35 Its July 2026 update reports that published families reached over 37,800 in 2025, and that generative AI’s share of all AI-related patent families rose from 4.2 per cent in 2017 to 8.7 per cent in 2025.36

On concentration, the 2024 report found that the top ten applicants for generative AI patents over 2014–2023 were all large corporations or a national academy: Tencent, Ping An Insurance, Baidu, the Chinese Academy of Sciences, IBM, Alibaba, Samsung, Alphabet, ByteDance and Microsoft.37 The 2026 update notes that the field is increasingly led by large multinational enterprises across many industries.38 Inventor locations are similarly concentrated: in 2014–2023, China accounted for 38,210 generative AI inventions, against 6,276 for the United States, 4,155 for the Republic of Korea, 3,409 for Japan and 1,350 for India.39

Concentration is not unlawful, and patents reward real investment. But General Comment No. 25 warns that emerging inequalities are strongly linked to the capacity of some businesses to access, store and exploit massive data.40 When the same firms hold the data, the computing infrastructure and the patents, the patent system can reinforce a structural imbalance in who participates in, and who benefits from, AI-driven science.

B. The Indian legal framework

The Patents Act, 1970 was enacted with an explicit public-interest orientation. Its framework relevant to AI can be grouped under four heads:41

•  What can be patented. Section 2(1)(j) defines “invention”; Section 3 lists what are not inventions. Section 3(k) excludes “a mathematical or business method or a computer programme per se or algorithms”. Section 3(b) excludes an invention the primary or intended use or commercial exploitation of which could be contrary to public order or morality, or which causes serious prejudice to human, animal or plant life or health or to the environment.

•  Who may apply. Section 6 permits applications by the “true and first inventor” or assignee; Section 2(1)(y) defines that expression negatively.

•  What must be disclosed. Section 10(4) requires a complete specification to describe the invention fully and particularly, including the best method of performing it.

•  How the public interest is protected. Section 47 makes grants subject to conditions, including use for experiment or research and the instruction of pupils. Section 83 sets out general principles for the working of patents. Sections 84 and 92 provide for compulsory licences, Section 100 for government use, and Section 107A for certain non-infringing uses.

On subject matter, the Delhi High Court in Ferid Allani (2019) and later decisions held that computer related inventions showing a technical effect or technical contribution are not excluded merely because they use software.42 The CGPDTM’s CRI Guidelines, 2025 consolidate that jurisprudence. They introduce a step-wise test for Section 3(k), a dedicated chapter on AI, machine learning and deep learning with worked examples, and specific disclosure expectations for AI inventions, such as details of training data, model architecture and learning methods sufficient for a skilled person to reproduce the invention.43 The Guidelines also distinguish AI-generated from AI-assisted inventions and treat the former as unpatentable.44

On inventorship, the Indian Patent Office refused Application No. 202017019068 on 15 April 2026 under Section 15, holding that DABUS cannot be designated the “true and first inventor” under Sections 2(1)(y), 6, 7 and 10, and that the claims also lacked an inventive step.45 This aligns India with the United Kingdom, the United States, Germany, Japan and the European Patent Office.

C. Why this framework has not been read through a rights lens

The Indian framework thus contains both incentive mechanisms and public-interest safeguards. Yet discussion of AI patents in India has focused almost entirely on the former: how to draft claims that clear Section 3(k), and who may be named as inventor. The safeguards in Sections 3(b), 47, 83, 84, 92 and 100 have been developed largely in the context of pharmaceuticals and have rarely been considered for AI; neither the Patent Office’s guidance on AI inventions nor the Standing Committee’s chapter on AI addresses them.46 The Parliamentary Standing Committee on Commerce’s 161st Report (2021) called for the Patents Act, including Section 3(k), to be reviewed to accommodate AI, but it framed the issue mainly as one of encouraging AI innovation.47 The next two Parts examine what is lost by that narrow framing.

How AI patents advance the right to science

A human rights critique of patents must begin by acknowledging what patents do well. General Comment No. 25 itself recognises that intellectual property enhances science and technology through economic incentives and that private research can be instrumental to the right.48 In the AI context, patents can advance the right to science in at least four ways.

A. Incentive and investment (availability)

Developing AI systems for socially important uses, such as medical imaging, drug discovery or agricultural forecasting, requires large investments in data, computing and expertise. Patents allow developers to recover that investment and attract capital. To the extent that they bring forward AI applications that would not otherwise exist, patents contribute to the availability element of the right, which requires that scientific progress actually takes place.49

B. Disclosure as diffusion (availability and quality)

The patent bargain requires disclosure. In exchange for a limited monopoly, the inventor must describe the invention so that others can learn from it and, after expiry, use it. For AI, this function is particularly valuable. Much AI development takes place behind trade secrecy. A patent specification that sets out a model’s architecture, training approach and functioning places knowledge in the public domain that might otherwise remain hidden.

The CRI Guidelines, 2025 strengthen this function by expecting AI patent applications to disclose details such as the nature of training data, the model architecture and the learning methods, so that a person skilled in the art can reproduce the invention.50 Read through the right to science, this is not merely a technical requirement. It is a mechanism of diffusion, one of the duties under Article 15(2) ICESCR, and it supports the quality element by making AI claims verifiable.51

Patent documents can thus inform public debate about the risks of AI, supporting the participatory and precautionary dimensions of the right.52

C. Applications that realise other rights (accessibility)

The right to science is also instrumental: it is a means of realising other rights, particularly health.53 Many patented AI applications serve exactly that purpose. AI tools for early diagnosis, for screening of tuberculosis or cancer, and for accelerating drug discovery can extend the reach of health systems.

Assistive technologies offer a striking example. WIPO’s Technology Trends report on assistive technology found that over one billion people need such technologies, a figure expected to double in the next decade as populations age, and that only one in ten people currently have access to the assistive products they need.54 The report found that filings in emerging assistive technologies, many of which use AI, were growing about three times faster than in conventional assistive technology.55 General Comment No. 25 asks States to promote technologies that facilitate access to scientific education and employment for persons with disabilities, and to implement universal design.56 Where patents stimulate such technologies, they serve the right to science directly.

D. Rewarding human creators (Article 15(1)(c))

Finally, the patent system’s insistence on human inventorship, now confirmed in India by the DABUS refusal,57 is consistent with the human rights framework. Article 15(1)(c) ICESCR protects the moral and material interests of human authors of scientific productions.58 Requiring that a natural person be named as inventor ensures that human scientists who use AI as a tool continue to be credited. General Comment No. 25 also asks States to ensure that contributions to scientific knowledge are appropriately credited.59

These benefits, however, are not automatic. Each depends on how the patent system is designed and administered. The next Part examines the ways in which AI patents can work against the right.

How AI patents threaten the right to science

The three harms identified in General Comment No. 25, namely distortion of research priorities, restrictions on knowledge sharing and barriers to access,60 each take a distinctive form in AI. To these may be added two problems specific to AI: thin disclosure of opaque models, and patents over applications that themselves threaten rights.

A. Distortion of research priorities

Patent incentives reward inventions that can be sold at a profit. AI research is therefore drawn towards finance, advertising, consumer platforms and enterprise services, where returns are high. WIPO’s data on generative AI show patenting spreading into application areas such as security, telecommunications, and banking and finance, and a growing role for large enterprises across these sectors.61 By contrast, AI applications for neglected diseases, small-language education tools or rural public services may attract little private investment, because their users cannot pay.

This is precisely the distortion General Comment No. 25 describes with reference to neglected diseases.62 It does not mean patents should be abolished. It means that patent incentives alone cannot be relied on to direct AI towards the needs that the right to science prioritises, particularly the needs of persons living in poverty.63

B. Concentration and restrictions on follow-on research

AI research is cumulative. New models build on existing architectures, training methods and datasets. Where foundational techniques are covered by dense clusters of patents held by a few firms, smaller developers, universities and researchers in developing countries face legal uncertainty and licensing costs. This affects the participation dimension of the right, which protects everyone’s open opportunity to take part in scientific progress.64

The concentration described in Part III compounds the problem. Patents interact with control over data and computing. A researcher may be free to read a patent specification but unable to use the method without access to comparable data or infrastructure. General Comment No. 25’s call to regulate the ownership and control of data according to human rights principles reflects this concern.65

In India, Section 47(3) allows the use of a patented invention for the purpose merely of experiment or research, including the imparting of instructions to pupils.66 But the scope of this exemption has not been tested for AI, where “research” may involve training or fine-tuning a model on patented methods and then deploying the results. Uncertainty about its scope may itself deter research.

C. Barriers to access to essential applications

The third harm is pricing and exclusion. As AI becomes embedded in diagnostic devices, clinical decision support, assistive technologies and educational tools, patents may allow holders to set prices that exclude low-income users and public institutions. A 2025 study mapping AI, intellectual property and human rights in European health systems found tensions between promoting innovation and ensuring accessibility, quality and equity.67

AI also presents a particular access problem. A patent may cover a diagnostic method, but effective use may require the trained model, its parameters and validated data, which are typically held as trade secrets. Traditional access tools, such as a compulsory licence over the patent, may then be insufficient on their own, because the licensee obtains the right to use the method but not the know-how needed to make it work.

D. Thin disclosure of opaque models

The disclosure benefit described in Part IV is real only if specifications actually enable the invention. Many AI systems are difficult to describe in a way that permits reproduction. Applicants may claim broad functional results while describing the model and its training only in general terms. If patents are granted on such disclosures, the public pays the price of exclusivity without receiving the knowledge that justifies it. This undermines both the availability and quality elements of the right.68

The CRI Guidelines, 2025 recognise this risk by setting AI-specific disclosure expectations.69 Their effectiveness, however, depends on consistent application by examiners, and the Guidelines do not have the force of statute.

E. Patents over rights-threatening applications

Finally, some AI inventions are themselves capable of threatening rights: systems for mass surveillance, emotion recognition, biometric profiling or discriminatory scoring. General Comment No. 25 notes that algorithms can reinforce discrimination70 and asks that autonomous systems be designed to avoid discrimination and allow explanation and accountability.71 Iga Bałos has similarly observed that while some AI patents simply facilitate everyday life, others may pose a threat to human rights.72

A patent does not authorise the use of an invention; it only grants a right to exclude others. Indian courts have stressed this point in interpreting Section 3(b).73 Nonetheless, patents can confer commercial value and legitimacy on rights-threatening technologies. In a 2024 study, Alina Wernick concludes that excluding algorithmic surveillance patents on ordre public grounds is generally undesirable, because the patent system is ill-equipped to evaluate the impacts of exploitation, and because disclosure has positive externalities.74 This counter-argument deserves weight. It suggests that the right to science is better served by targeted safeguards and transparency than by broad exclusions.

F. Summary

Table 2 summarises how each concern identified in General Comment No. 25 takes shape for AI patents, together with the two problems specific to AI.

General Comment No. 25 concern AI-specific form Element of the right affected
Distortion of research funding AI investment flows to profitable sectors; neglect of low-income needs Availability, accessibility
Restrictions on knowledge sharing Patent clusters plus control of data and compute; uncertain research exemption Participation, availability
Barriers to access High prices for AI diagnostics and assistive tools; patent plus trade secret Accessibility
(AI-specific) Thin disclosure Broad functional claims over opaque models Quality, availability
(AI-specific) Rights-threatening applications Surveillance, profiling, discriminatory systems Acceptability

Table 2: How AI patents can threaten the right to science

Reading the right to science into Indian patent law

India has not enacted a statute on the right to science, and the Patents Act, 1970 does not mention it. This Part argues that neither is necessary for the right to begin shaping AI patent practice. The Constitution provides the interpretive basis, and the Act already contains the instruments.

A. Constitutional basis

Three constitutional provisions support reading the right to science into patent law:75

•  Article 21. The Supreme Court has read the right to life expansively to include the right to health and, in K.S. Puttaswamy v. Union of India, the right to privacy.76 Access to the applications of science that are essential to health and dignity falls naturally within this protection.

•  Article 51(c). The State shall endeavour to foster respect for international law and treaty obligations. In Vishaka v. State of Rajasthan, the Supreme Court held that international conventions consistent with fundamental rights, and not inconsistent with domestic law, may be read into those rights.77 India’s obligations under Article 15 ICESCR can therefore inform the interpretation of statutes such as the Patents Act.

•  Article 51A(h). It is a fundamental duty of every citizen to develop the scientific temper, humanism and the spirit of inquiry and reform. This duty resonates with the participatory conception of the right to science in General Comment No. 25.78

The Supreme Court’s decision in Novartis AG v. Union of India demonstrates that Indian courts will interpret the Patents Act purposively, with regard to its legislative history and public-interest objectives.79 The Court’s reading of Section 3(d) gave effect to the legislature’s concern to prevent the extension of monopolies without genuine therapeutic advance. A similar purposive approach can be taken to AI.

B. Section 83: the Act’s own statement of the right to science

Section 83 lays down general considerations for the working of patents. Among them, patents are granted to encourage inventions and to secure that they are worked in India on a commercial scale; the protection of patent rights should contribute to the promotion of technological innovation and to the transfer and dissemination of technology, to the mutual advantage of producers and users of technological knowledge and in a manner conducive to social and economic welfare; patents should not impede the protection of public health and should act as an instrument to promote public interest in sectors of vital importance; and patents should make the benefit of the invention available at reasonably affordable prices to the public.80

These principles closely track the elements of the right to science: dissemination (availability), affordability (accessibility) and social welfare. Section 83 can therefore serve as the statutory bridge through which General Comment No. 25 is applied to AI. It is directly relevant when the Controller considers compulsory licence applications, and it can guide interpretation of other provisions.

C. Mapping the right to science onto the Patents Act

Table 3 maps each element of the right to science onto the provisions of the Patents Act, 1970 that can give effect to it in the AI context.81

Right to science element Provision of the Patents Act, 1970 Application to AI
Availability and diffusion Section 10(4): full and particular description; best method Insist on reproducible disclosure of AI models as set out in the CRI Guidelines, 2025; refuse or limit broad functional claims that are not enabled
Quality Sections 10(4), 25 and 64 (insufficiency as a ground of opposition and revocation) Use pre- and post-grant opposition to challenge AI patents with inadequate disclosure
Participation and freedom of research Section 47(3): use for experiment, research and instruction Clarify that training, testing and studying AI models for research falls within the exemption
Accessibility Sections 83, 84, 92 and 100 Compulsory licences and government use for essential AI applications in health, disability and education
Acceptability and protection from harm Section 3(b): public order, morality, serious prejudice to human life or health Narrow, intent-based scrutiny of AI inventions whose primary purpose is unlawful discrimination or surveillance
Balanced subject matter Section 3(k): computer programme per se, algorithms Keep abstract algorithms and models in the public domain; grant patents only for technical contributions
Credit to human creators Sections 2(1)(y), 6 and 10; DABUS refusal Maintain human inventorship while requiring honest disclosure of the role of AI

Table 3: The right to science mapped onto the Patents Act, 1970

D. Section 3(k) as a guardian of the scientific commons

Section 3(k) is often presented as an obstacle to innovation. From the right to science perspective, it has a protective function. By excluding algorithms and computer programmes “per se”, it keeps the basic building blocks of AI, mathematical methods and abstract learning techniques, in the public domain where all researchers can use them. The technical contribution test developed by Indian courts and codified in the CRI Guidelines, 2025 allows protection for genuine technical applications while preserving that commons.82

This balance is worth preserving. Calls to relax Section 3(k), such as those in the 161st Report of the Parliamentary Standing Committee,83 should be assessed not only for their effect on investment but for their effect on the availability of basic AI knowledge to Indian researchers. The UK’s adoption of a low “any hardware” threshold in Emotional Perception84 should be studied with this in mind; a lower threshold at the eligibility stage shifts more of the burden of protecting the commons onto inventive step and disclosure.

E. Section 3(b) and acceptability

Section 3(b) excludes an invention the primary or intended use or commercial exploitation of which could be contrary to public order or morality, or which causes serious prejudice to human, animal or plant life or health or to the environment. The Calcutta High Court has held that this is an “intent” principle rather than an “effect” principle, and that the grant of a patent does not by itself confer a right to commercialise.85

These limits are sound. Section 3(b) cannot be used to refuse general-purpose AI inventions merely because they could be misused. It may, however, be relevant to the rare AI invention whose stated and primary purpose is, for example, covert mass surveillance of individuals or unlawful discrimination on protected grounds. Reading Section 3(b) in light of Articles 14, 15 and 21 of the Constitution and General Comment No. 25’s concern with discriminatory algorithms86 would give examiners a principled basis for such rare cases, while respecting the counter-argument that patent offices are poorly placed to regulate technology.87

F. Compulsory licensing and government use

India has used compulsory licensing to protect access. In Natco Pharma Ltd. v. Bayer Corporation, the Controller granted India’s first compulsory licence under Section 84 for the anti-cancer drug sorafenib on the grounds that the reasonable requirements of the public were not met, the drug was not available at a reasonably affordable price and the patent was not worked in India; the licence was upheld by the Intellectual Property Appellate Board and the Bombay High Court.88 The same provisions apply to any patented invention, including AI.

Two adaptations are needed for AI. First, the “reasonable requirements of the public” under Section 84 should be interpreted, in light of Section 83 and General Comment No. 25, to include access to AI applications essential for health, disability and education. Second, because AI patents are often accompanied by trade secrets over trained models and data, the effectiveness of a licence may depend on the terms on which know-how is transferred. Government use under Section 100, which permits use of inventions for the purposes of government, may be especially useful for public health and public education deployments of AI.

G. Inventorship and the human scientist

India’s DABUS refusal89 is consistent with the right to science. Article 15(1)(c) ICESCR protects human authors, not machines.90 Insisting on a natural-person inventor keeps human scientists at the centre of the patent system. However, the right to science also requires honesty about the role of AI. If inventions largely produced by AI are attributed to humans who contributed little, the patent system rewards control over AI tools rather than human scientific contribution. The CRI Guidelines’ distinction between AI-generated and AI-assisted inventions91 needs a clearer threshold of human contribution, together with a duty to disclose the role of AI in the inventive process.

Recommendations

The following recommendations are framed so that most can be implemented through examination practice, guidelines and judicial interpretation, without immediate legislative amendment.

A. For the Patent Office (CGPDTM)

1.  Enforce AI disclosure standards consistently. The disclosure expectations in the CRI Guidelines, 2025 should be applied rigorously, and broad functional claims over AI models that are not enabled by the specification should be refused or narrowed. Disclosure should be treated as a public-interest obligation, not a formality.

2.  Require disclosure of the role of AI. Applicants should state whether and how AI was used in making the claimed invention, so that the AI-generated and AI-assisted distinction in the Guidelines can be applied in practice.

3.  Issue guidance on Section 3(b) for AI. Guidance should confirm that Section 3(b) applies only on an intent basis, while identifying the narrow categories of AI inventions whose primary purpose would be contrary to public order, such as covert mass surveillance or unlawful discrimination.

B. For the Government (DPIIT)

1.  Clarify the research exemption. Section 47(3) should be clarified, by rules, guidelines or amendment, to confirm that research uses of patented AI methods, including training and evaluation of models for non-commercial research and teaching, are permitted.

2.  Preserve the Section 3(k) balance. Any review of Section 3(k) should assess its effect on access to basic AI knowledge, not only on investment, and should retain the exclusion of algorithms and models in the abstract.

3.  Prepare access mechanisms for essential AI. The Government should identify AI applications essential for health, disability access and education, and be prepared to use Sections 84, 92 and 100 where access is unreasonably restricted, with terms addressing transfer of necessary know-how.

4.  Align funding with need. Consistent with General Comment No. 25, public funding, procurement and prize mechanisms should support AI for neglected needs, so that research direction is not left to patent incentives alone.92

5.  Adopt a right-to-science framework. General Comment No. 25 asks States to adopt a national framework and action plan that harmonise intellectual property with the right.93 India’s AI policy documents should expressly integrate this objective.

C. For courts and tribunals

1.  Interpret the Patents Act in light of Section 83 and Article 15 ICESCR. Following Novartis and Vishaka, courts should read Sections 3(k), 3(b), 10, 47 and 84 purposively, with regard to the right to enjoy the benefits of scientific progress.94

D. For the research community

1.  Use opposition proceedings. Universities and public research institutions should use pre-grant and post-grant opposition to challenge over-broad or insufficiently disclosed AI patents, as public interest groups have done in the pharmaceutical field, among them the Cancer Patients Aid Association in its pre-grant opposition to Novartis’s application for the beta crystalline form of imatinib mesylate.95

Conclusion

The legal debate on AI and patents has asked whether machines can invent and whether algorithms can be owned. These are important questions, and a broad international consensus has now formed around them: AI is a tool, inventors must be human, and only technical contributions are patentable. India’s DABUS refusal and the CRI Guidelines, 2025 place India firmly within that consensus.

This article has argued that the more consequential question is distributive. AI patents can advance the right to science by drawing investment into useful applications, by placing knowledge about AI in the public domain and by stimulating technologies that serve health and disability access. They can also undermine it by steering research away from the needs of the poor, by entrenching concentrated control over AI knowledge and infrastructure, by pricing essential applications out of reach, by rewarding thin disclosure of opaque models, and by lending value to applications that threaten rights.

General Comment No. 25 does not ask States to choose between innovation and access. It asks them to foster the positive effects of intellectual property while avoiding its negative effects, and to treat intellectual property as a social product with a social function.96 Indian law is unusually well placed to answer that call. Its Constitution invites the reading of international obligations into domestic rights; its courts have shown in Novartis that the Patents Act will be read purposively; and the Act itself, through Sections 3(b), 3(k), 10, 47, 83, 84, 92 and 100, already contains the instruments needed.

What has been missing is the perspective that connects them. If India’s patent system is to serve not only those who own AI but everyone who is entitled to share in its benefits, the right to science must become part of how AI patents are examined, granted, licensed and litigated.

*****

Footnotes

1. World Intellectual Property Organization, Patent Landscape Report: Generative Artificial Intelligence (GenAI) (2024), https://www.wipo.int/web-publications/patent-landscape-report-generative-artificial-intelligence-genai/en/key-findings-and-insights.html [hereinafter WIPO GenAI Report]; WIPO, China-Based Inventors Filing Most GenAI Patents, WIPO Data Shows, PR/2024/921 (July 3, 2024), https://www.wipo.int/pressroom/en/articles/2024/article_0009.html (last visited Nov. 10, 2026) [hereinafter WIPO 2024 Press Release].

2. WIPO, GenAI Innovation Soaring, With Patent Activity Nearly Tripling in Two Years, PR/2026/957 (July 14, 2026), https://www.wipo.int/pressroom/en/articles/2026/article_0012.html (last visited Nov. 10, 2026) [hereinafter WIPO 2026 Update].

3. WIPO 2024 Press Release, supra note 1.

4. Controller of Patents & Designs, Order Refusing Patent Application No. 202017019068 (Applicant Stephen L. Thaler; Inventor Named as DABUS) (Indian Patent Office Apr. 15, 2026) (India), https://www.bananaip.com/intellepedia/wp-content/uploads/2026/04/DABUS-Patent-Application-Refusal-India-AI-Inventorship-Order.pdf [hereinafter DABUS Refusal Order].

5. Revised Inventorship Guidance for AI-Assisted Inventions, 90 Fed. Reg. 54,636 (Nov. 28, 2025).

6. The Patents Act, 1970, No. 39, Acts of Parliament, 1970 (India) [hereinafter Patents Act]; Office of the Controller General of Patents, Designs & Trade Marks, Guidelines for Examination of Computer Related Inventions (CRIs), 2025 (July 29, 2025) [hereinafter CRI Guidelines, 2025]; Ferid Allani v. Union of India, W.P.(C) 7/2014 (Del. HC Dec. 12, 2019) (India); Microsoft Tech. Licensing LLC v. Assistant Controller of Patents & Designs, C.A.(COMM.IPD-PAT) 185/2022 (Del. HC Apr. 16, 2024) (India).

7. Emotional Perception AI Ltd. v. Comptroller-General of Patents, Designs & Trade Marks [2026] UKSC 3 (appeal taken from Eng.).

8. G.A. Res. 217 (III) A, Universal Declaration of Human Rights, art. 27 (Dec. 10, 1948) [hereinafter UDHR]; International Covenant on Economic, Social and Cultural Rights art. 15, Dec. 16, 1966, 993 U.N.T.S. 3 [hereinafter ICESCR].

9. Comm. on Econ., Soc. & Cultural Rts., General Comment No. 25 (2020) on Science and Economic, Social and Cultural Rights (Article 15 (1) (b), (2), (3) and (4) of the International Covenant on Economic, Social and Cultural Rights), U.N. Doc. E/C.12/GC/25 (Apr. 30, 2020), https://digitallibrary.un.org/record/3899847/files/E_C.12_GC_25-EN.pdf [hereinafter General Comment No. 25].

10. Id. ¶¶ 60–62, 72–76.

11. UDHR, supra note 8, art. 27.

12. ICESCR, supra note 8, art. 15.

13. ICESCR, supra note 8; Status of Treaties: International Covenant on Economic, Social and Cultural Rights, United Nations Treaty Collection, https://treaties.un.org/Pages/ViewDetails.aspx?src=TREATY&mtdsg_no=IV-3&chapter=4 (last visited Nov. 19, 2026) (recording India’s accession on Apr. 10, 1979).

14. General Comment No. 25, supra note 9, ¶ 2.

15. Id. ¶ 8.

16. Id. ¶¶ 9–11.

17. Id. ¶¶ 16–20.

18. Comm. on Econ., Soc. & Cultural Rts., General Comment No. 17 (2005): The Right of Everyone to Benefit from the Protection of the Moral and Material Interests Resulting from Any Scientific, Literary or Artistic Production of Which He or She Is the Author, ¶ 1, U.N. Doc. E/C.12/GC/17 (Jan. 12, 2006) [hereinafter General Comment No. 17].

19. General Comment No. 25, supra note 9, ¶ 12.

20. Farida Shaheed (Special Rapporteur in the Field of Cultural Rights), Report of the Special Rapporteur in the Field of Cultural Rights: Patent Policy and the Right to Science and Culture, ¶ 5, U.N. Doc. A/70/279 (Aug. 4, 2015), https://digitallibrary.un.org/record/801962/files/A_70_279-EN.pdf.

21. Id. at 2 (summary).

22. General Comment No. 25, supra note 9, ¶ 60.

23. Id. ¶ 61.

24. Id. ¶ 62.

25. Id.

26. Id. ¶ 69.

27. Id. ¶ 72.

28. Id. ¶ 73.

29. Id. ¶¶ 74–76.

30. Id. ¶ 41.

31. Id. ¶ 43.

32. Id. ¶ 52.

33. Id. ¶ 86.

34. Id. ¶ 89.

35. WIPO GenAI Report, supra note 1.

36. WIPO 2026 Update, supra note 2.

37. WIPO GenAI Report, supra note 1; WIPO 2024 Press Release, supra note 1.

38. WIPO 2026 Update, supra note 2.

39. WIPO 2024 Press Release, supra note 1.

40. General Comment No. 25, supra note 9, ¶ 76.

41. Patents Act, supra note 6, §§ 2(1)(j), 2(1)(y), 3, 6, 10(4), 47, 83, 84, 92, 100, 107A.

42. Ferid Allani, supra note 6; Microsoft Tech. Licensing, supra note 6.

43. CRI Guidelines, 2025, supra note 6, §§ 4.5, 5.2, 5.2.2–5.2.3.

44. Id. § 5.2.1.

45. DABUS Refusal Order, supra note 4, ¶ 36.

46. See CRI Guidelines, 2025, supra note 6, ch. 5 (addressing AI inventions under Sections 3(k), 6 and 10, but none of these safeguards); Department-Related Parliamentary Standing Committee on Commerce, Rajya Sabha, One Hundred Sixty-First Report: Review of the Intellectual Property Rights Regime in India ¶¶ 8.1–8.7 (presented July 23, 2021) [hereinafter 161st Report] (chapter on artificial intelligence and IPR, likewise confined to subject matter and inventorship).

47. 161st Report, supra note 46, ¶¶ 8.2–8.7.

48. General Comment No. 25, supra note 9, ¶¶ 58, 60.

49. Id. ¶ 16.

50. CRI Guidelines, 2025, supra note 6, §§ 5.2.2–5.2.3.

51. General Comment No. 25, supra note 9, ¶¶ 14, 18.

52. Id. ¶¶ 56–57.

53. Id. ¶ 63.

54. World Intellectual Property Organization, WIPO Technology Trends 2021: Assistive Technology (2021); WIPO, WIPO Report Finds Significant Growth in Assistive Technologies as They Find Greater Application in Consumer Goods, PR/2021/875 (Mar. 23, 2021), https://www.wipo.int/pressroom/en/articles/2021/article_0003.html.

55. WIPO Technology Trends 2021, supra note 54.

56. General Comment No. 25, supra note 9, ¶ 35.

57. DABUS Refusal Order, supra note 4, ¶ 36.

58. ICESCR, supra note 8, art. 15(1)(c); General Comment No. 17, supra note 18, ¶ 1.

59. General Comment No. 25, supra note 9, ¶ 83.

60. Id. ¶ 61.

61. WIPO GenAI Report, supra note 1; WIPO 2026 Update, supra note 2.

62. General Comment No. 25, supra note 9, ¶ 61.

63. Id. ¶ 37.

64. Id. ¶¶ 10, 17.

65. Id. ¶ 76.

66. Patents Act, supra note 6, § 47(3).

67. Robin van Kessel et al., Artificial Intelligence, Intellectual Property, and Human Rights: Mapping the Legal Landscape in European Health Systems, 2 npj Health Sys. 43 (2025), https://doi.org/10.1038/s44401-025-00050-3.

68. General Comment No. 25, supra note 9, ¶¶ 16, 18.

69. CRI Guidelines, 2025, supra note 6, §§ 5.1, 5.2.2.

70. General Comment No. 25, supra note 9, ¶ 73.

71. Id. ¶ 75.

72. Iga Bałos, The Real Problems with Patents for AI-Related Inventions and Human Rights, in Artificial Intelligence and Human Rights 440, 440 (Dykinson 2021).

73. ITC Ltd. v. Controller of Patents, Designs & Trademark, IPDPTA No. 13 of 2024, ¶ 13 (Cal. HC May 20, 2025) (India).

74. Alina Wernick, Ordre Public Exceptions for Algorithmic Surveillance Patents, in Kreation Innovation Märkte – Creation Innovation Markets: Festschrift Reto M. Hilty 469 (Florent Thouvenin, Alexander Peukert, Thomas Jaeger & Christophe Geiger eds., 2024), https://doi.org/10.1007/978-3-662-68599-0_33.

75. India Const. arts. 21, 51(c), 51A(h).

76. Paschim Banga Khet Mazdoor Samity v. State of West Bengal, (1996) 4 SCC 37 (India); Justice K.S. Puttaswamy (Retd.) v. Union of India, (2017) 10 SCC 1 (India).

77. Vishaka v. State of Rajasthan, (1997) 6 SCC 241 (India).

78. General Comment No. 25, supra note 9, ¶¶ 9–11.

79. Novartis AG v. Union of India, (2013) 6 SCC 1 (India).

80. Patents Act, supra note 6, § 83.

81. Patents Act, supra note 6, §§ 2(1)(y), 3(b), 3(k), 6, 10, 25, 47(3), 64, 83, 84, 92, 100; CRI Guidelines, 2025, supra note 6, § 5.2.2; DABUS Refusal Order, supra note 4, ¶ 36.

82. CRI Guidelines, 2025, supra note 6, § 4.5; Ferid Allani, supra note 6.

83. 161st Report, supra note 46, ¶¶ 8.3, 8.5–8.7.

84. Emotional Perception, supra note 7.

85. ITC Ltd. v. Controller of Patents, Designs & Trademark, IPDPTA No. 121 of 2023, ¶ 37 (Cal. HC Apr. 30, 2025) (India); ITC Ltd., IPDPTA No. 13 of 2024, supra note 73, ¶¶ 11, 13.

86. General Comment No. 25, supra note 9, ¶¶ 73, 75.

87. Wernick, supra note 74.

88. Natco Pharma Ltd. v. Bayer Corp., Compulsory Licence Application No. 1 of 2011 (Controller of Patents Mar. 9, 2012) (India), aff’d, Bayer Corp. v. Union of India (IPAB Mar. 4, 2013) (India), writ petition dismissed, Bayer Corp. v. Union of India, Writ Petition No. 1323 of 2013 (Bom. HC July 15, 2014) (India).

89. DABUS Refusal Order, supra note 4, ¶ 36.

90. ICESCR, supra note 8, art. 15(1)(c); General Comment No. 17, supra note 18, ¶ 1.

91. CRI Guidelines, 2025, supra note 6, § 5.2.1.

92. General Comment No. 25, supra note 9, ¶ 62.

93. Id. ¶¶ 52, 86.

94. Novartis, supra note 79; Vishaka, supra note 77.

95. Novartis, supra note 79, ¶ 13 & n.6 (recording the five pre-grant oppositions to the application, including that of the Cancer Patients Aid Association).

96. General Comment No. 25, supra note 9, ¶ 62.

How to cite

Laxmi Walia & Dr Namrata Yadav, Who Benefits from Machine Invention? The Right to Science, Artificial Intelligence and the Indian Patent System, 9 Int'l J.L. Mgmt. & Human. 1146 (2026), https://doi.org/10.63108/IJLMH.13301.

Cite this article

Laxmi Walia & Dr Namrata Yadav, Who Benefits from Machine Invention? The Right to Science, Artificial Intelligence and the Indian Patent System, 9 Int'l J.L. Mgmt. & Human. 1146 (2026), https://doi.org/10.63108/IJLMH.13301.
Chat with us