abstract
The rapid advancement of artificial intelligence has precipitated a fundamental challenge to traditional patent law frameworks, particularly regarding the inventor requirement. This research article examines the legal status of AI-generated inventions under Indian patent law, focusing on whether AI systems can be recognized as inventors under the Patents Act, 1970. Through doctrinal analysis and comparative examination of international jurisprudence, this paper explores statutory provisions, judicial interpretations, and administrative positions adopted by the Indian Patent Office. The research analyses landmark DABUS applications across multiple jurisdictions and their implications for Indian patent jurisprudence. The study reveals that current Indian patent law, premised on human inventorship, inadequately addresses AI-generated inventions. The article concludes with recommendations for reforming the Patents Act to accommodate AI-generated inventions while preserving policy objectives including disclosure requirements and accountability mechanisms. The research employs black letter methodology, analysing statutory text, case law, and comparative international approaches to provide a comprehensive assessment of this emerging legal challenge.
keywords
Artificial Intelligence, Patent Law, Inventorship, DABUS, Patents Act 1970, AI-Generated Inventions
introduction
The intersection of artificial intelligence and patent law represents one of the most consequential legal challenges of the twenty-first century. As AI systems evolve from mere tools to autonomous systems capable of generating novel inventions, traditional patent law frameworks face unprecedented challenges. The Indian patent system, governed by the Patents Act, 1970, was designed when human agency was the unquestioned prerequisite for inventive activity. Today, machine learning algorithms and neural networks produce innovations across pharmaceutical research, materials science, and engineering, often with minimal human intervention.
The fundamental question confronting Indian patent law is whether inventions generated by AI systems satisfy the statutory requirement of inventorship. Section 2(1)(y) of the Patents Act defines “true and first inventor” as one who does not derive the invention from any other person, while Section 6 specifies who may apply for patents.[1] These provisions implicitly assume human inventorship, creating a categorical barrier for AI-generated inventions.
The DABUS (Device for the Autonomous Bootstrapping of Unified Sentience) cases have catalyzed global debate on AI inventorship. Dr. Stephen Thaler’s applications designating DABUS, an AI system, as the sole inventor have been adjudicated in multiple jurisdictions with divergent outcomes.[2] While South Africa initially granted patents recognizing AI inventorship, courts in the United Kingdom, United States, and European Patent Office rejected such applications.[3] India received similar DABUS applications in 2020-2021, and the Controller General of Patents rejected them, maintaining that only natural persons qualify as inventors under Indian law.
This resistance to AI inventorship raises critical policy questions. If AI-generated inventions are denied patent protection due to the absence of human inventors, such innovations enter the public domain immediately, potentially undermining investment incentives. Conversely, recognizing AI systems as inventors challenges foundational principles including the disclosure bargain, moral rights of inventors, and accountability mechanisms.
The Indian context presents unique considerations. India’s patent regime, shaped by developmental objectives and public health concerns, incorporates stringent patentability standards including enhanced inventive step requirements under Section 3.[4] The judiciary has emphasized substance over form, as evidenced in Novartis AG v. Union of India.[5] Extending this approach to AI-generated inventions requires careful analysis of whether such inventions satisfy novelty, inventive step, and industrial applicability requirements.
This research undertakes a comprehensive examination of AI-generated inventions within the Indian patent law framework, analysing statutory provisions, administrative practices, and judicial precedents. Through comparative analysis of international approaches, the paper identifies best practices and proposes pragmatic solutions balancing innovation incentives with legal certainty and policy coherence.
research methodology
This research employs doctrinal legal research methodology, the primary approach for analysing statutory provisions, case law, and legal principles governing patent law. The doctrinal method involves systematic analysis of legal texts, including legislation, judicial decisions, administrative guidelines, and scholarly commentary, to determine the current state of law and identify gaps or ambiguities.
review of literature
Scholarly discourse on AI-generated inventions has proliferated since 2019 following the first DABUS applications. Traditional patent law scholarship has articulated multiple theoretical justifications: utilitarian incentive theory, natural rights theory, disclosure theory, and prospect theory.[6] These frameworks presuppose human inventorship.
Recent scholarship by Ryan Abbott, particularly through “The Artificial Inventor Project,” has challenged anthropocentric assumptions. Abbott argues patent law’s fundamental objective is promoting innovation, not rewarding human inventors per se.[7] From this instrumentalist perspective, denying patent protection to AI-generated inventions frustrates patent law’s core purpose.
Comparative legal scholarship has documented divergent national responses. The UK Supreme Court in Thaler v. Comptroller-General held that UK patent law unambiguously requires inventors to be natural persons.[8] The court emphasized that inventorship rights are intrinsically linked to derivation of title and that recognizing AI inventorship would create insurmountable practical difficulties.
The Australian Federal Court initially recognized DABUS as an inventor in Thaler v. Commissioner of Patents,[9] but the Full Federal Court subsequently overturned this decision,
and the High Court declined special leave to appeal, effectively rejecting AI inventorship.[10]
The EPO’s approach emphasizes that the European Patent Convention requires inventors to be designated by name, presupposing natural persons. The EPO Legal Board of Appeal in Decision J 8/20 concluded that designating a machine as inventor fails to satisfy formal requirements.[11]
Indian intellectual property scholarship on AI inventorship remains nascent. Scholars including Prashant Reddy have examined Indian patent law’s inventorship requirements, noting that the Patents Act’s definition implicitly requires human inventors.[12] Academic commentary addresses the Patent Office’s rejection of DABUS applications, though scholars note these decisions fail to engage substantively with policy implications.
analysis: the legal status of ai-generated inventions
Statutory Framework
Indian patent law does not explicitly address whether AI systems can be recognized as inventors. However, statutory analysis reveals that the Patents Act, 1970, implicitly requires inventors to be natural persons. Section 2(1)(s) defines “inventor” as “a person who actually devises the invention.”[13] The critical interpretative question concerns the meaning of “person.”
Under the General Clauses Act, 1897, “person” includes “any company or association or body of individuals, whether incorporated or not.”[14] However, this definition encompasses juridical persons, not autonomous machines or AI systems. Nothing in the Patents Act suggests Parliament intended to extend “inventor” status to non-human entities beyond juridical persons.
Section 2(1)(y) defines “true and first inventor” as excluding persons deriving inventions from others.[15] This provision establishes that inventors must be originators. The question arises whether AI systems “devise” inventions or simply execute computational processes. If AI systems lack intentionality, arguably they do not “devise” inventions in the contemplated sense.
Section 6 specifies who may apply for patents: persons claiming to be true and first inventors, their assignees, or legal representatives. This links patent applications to identified inventors. Rule 13 of the Patents Rules requires naming and addressing the inventor,[16] presupposing human inventors capable of identification by name and address.
Administrative Practice
The Indian Patent Office received DABUS applications in 2020-2021. The Controller rejected these applications, reasoning that Section 2(1)(s) requires inventors to be “persons,” meaning natural persons in legal context.[17] Patent applications must identify inventors by name and address, which cannot be satisfied by AI systems. The Controller concluded that recognizing AI inventorship would contradict the Patents Act’s structure linking inventorship to ownership rights and accountability.
Ownership and Accountability
Recognizing AI systems as inventors would create uncertainty regarding patent ownership. The Patents Act establishes that inventors or their assignees hold patent rights. If AI is the inventor, ownership becomes problematic as AI cannot hold property rights. Should ownership vest in the AI’s programmer, owner, or operator? Each possibility raises difficulties.
Additionally, patent law incorporates accountability mechanisms premised on human agency. Section 120 imposes criminal liability for false suggestions of patent rights.[18] These structures presuppose human decision-makers who can be held responsible. Allowing AI inventorship without clear human accountability could undermine patent system integrity.
Policy Implications
While statutory analysis indicates current law does not permit AI inventorship, policy analysis reveals concerns. If AI-generated inventions cannot be patented due to lack of human inventors, companies may reduce investment in AI research. Additionally, AI-generated inventions would enter the public domain immediately, eliminating exclusivity periods that fund innovation cycles.
Conversely, recognizing AI inventorship presents risks. Companies with sophisticated AI capabilities may generate vast patent applications, creating patent thickets impeding follow-on innovation. If AI-generated patents lack meaningful disclosure due to AI opacity, they may fail to serve patent law’s informational function.
Comparative Analysis
International divergence provides valuable insights. The UK approach requires legislative intervention for fundamental patent law changes, reflecting judicial modesty.[19] The US approach similarly delegates AI inventorship questions to Congress. The EPO’s procedural focus provides legal certainty while leaving policy reform to member states.[20]
For India, these approaches suggest fundamental reforms should originate from parliamentary action rather than judicial interpretation. The Patents Act’s text, structure, and history do not provide sufficient flexibility to recognize AI inventorship absent legislative amendment.
Doctrinal Accommodations
While recognizing AI as inventors appears incompatible with current law, certain approaches might accommodate AI-generated inventions:
- Tool Theory: AI systems could be characterized as tools used by human inventors, maintaining human inventorship while acknowledging AI’s role. However, this becomes tenuous when AI contributes the core inventive concept with minimal human intervention.[21] Under this approach, individuals who configure, train, or deploy AI systems could be named as inventors analogous to researchers using sophisticated laboratory equipment. Yet this analogy breaks down when examining the level of creative contribution. Traditional tools, however sophisticated, do not independently generate inventive concepts—they amplify or facilitate human creativity. Modern AI systems, particularly those employing deep learning and generative algorithms, may autonomously identify problems, explore solution spaces, and generate novel outputs that exceed their designers’ expectations or intentions. This autonomous creative capacity distinguishes AI from traditional tools and renders the tool theory analytically unsatisfying for truly AI-generated inventions.
- Joint Inventorship: Multiple individuals involved in developing or operating AI systems might be named as joint inventors. However, this requires genuine contribution to conception, risking false inventorship if those persons did not actually conceive the invention.[22] Indian patent law, following established jurisprudence in cases such as University of Rochester v. G.D. Searle & Co., requires that joint inventors each contribute to the conception of the claimed invention. Mere assistance in reducing an invention to practice, conducting routine testing, or following instructions does not constitute joint inventorship. Applying this standard to AI-generated inventions creates conceptual difficulties. If an AI system independently conceives the inventive concept while programmers, trainers, or operators provide infrastructure without contributing to the specific invention’s conception, naming these individuals as joint inventors would constitute false inventorship. This approach may pragmatically enable patent applications but compromises inventorship doctrine’s integrity.
- Employer Inventorship: Entities owning AI systems could claim patent rights with human employees named as inventors based on research contributions. This pragmatically enables patent protection but may distort inventorship requirements. Indian law recognizes that inventions made by employees in the course of employment may vest in employers through contractual assignment. However, this presupposes that employees are actual inventors. Simply naming employees as inventors for AI-generated inventions to satisfy formal requirements, when those employees did not genuinely conceive the inventions, amounts to false inventorship declarations subject to penalties under Section 120 of the Patents Act.
Each doctrinal accommodation involves stretching patent law principles beyond their conceptual limits and creates false inventorship risks. These compromises demonstrate that accommodating AI-generated inventions within existing legal frameworks produces unsatisfactory results, highlighting the urgent need for legislative clarity and reform specifically addressing this technological reality.
suggestions and recommendations
Addressing AI-generated inventions requires comprehensive legal reform. The following recommendations outline pathways for reform:
Legislative Amendment
Parliament should amend the Patents Act to explicitly address AI-generated inventions:
- Defining AI-Generated Inventions: The Act should define “AI-generated invention” as inventions where AI systems make substantial contributions without determinative human intervention, establishing examination standards. This definition should distinguish between AI-assisted inventions, where humans remain the primary creative agents using AI as a tool, and truly AI-generated inventions where the AI system autonomously generates the inventive concept. Clear definitional boundaries prevent uncertainty and ensure consistent application across patent examination processes. The definition should also specify technical thresholds for determining when AI contribution becomes “substantial,” potentially referencing factors such as the degree of human intervention in problem identification, solution generation, and selection of final embodiments.
- Establishing Inventorship Requirements: Amendments should specify that while AI systems may contribute to inventions, human persons or juridical entities must be designated as applicants with ownership rights, requiring disclosure of AI involvement. This approach separates inventorship from ownership, acknowledging AI’s creative contribution while vesting property rights in legally recognized persons. The amendment should establish a registry linking AI-generated patents to the AI systems involved, enabling tracking and accountability. Additionally, legislation should address whether AI developers, owners, operators, or other stakeholders have preferential ownership claims, providing clear priority rules that prevent ownership disputes.
- Enhanced Disclosure Obligations: Applications should include detailed descriptions of the AI system used, training data, human contributions, and the generation process, serving patent law’s informational objective. Enhanced disclosure requirements should mandate explanation of: (a) the AI architecture and learning methodology; (b) the nature and scope of training data; (c) the problem presented to the AI system; (d) the process by which the AI generated potential solutions; (e) criteria used to select the claimed invention from AI-generated alternatives; and (f) any human modifications to AI outputs. These disclosure requirements serve multiple functions: enabling skilled persons to reproduce the invention, facilitating patent examination by providing examiners sufficient information to assess patentability, and creating transparency regarding the extent of AI versus human contribution. However, disclosure requirements must balance transparency with protection of proprietary AI algorithms, potentially through confidential submissions to the Patent Office.
- Patentability Standards: The Act should clarify that AI-generated inventions must satisfy existing patentability criteria evaluated against prior art as known to persons skilled in the relevant field.[23] The inventive step assessment presents particular challenges for AI-generated inventions. Traditional inventive step analysis asks whether the invention would be obvious to a person skilled in the art having regard to the prior art. When AI systems analyse vast datasets and identify non-obvious patterns beyond human cognitive capacity, determining obviousness becomes conceptually complex. The legislation should clarify whether the “person skilled in the art” standard assumes a human equipped with AI tools or remains purely anthropocentric. Additionally, the industrial applicability requirement should be rigorously applied to ensure AI-generated inventions serve practical utility rather than generating patents for purely abstract or speculative concepts.
Institutional Reforms
The Indian Patent Office requires capacity building:
- Examiner Training: Patent examiners need training in AI technologies and machine learning to effectively examine AI-related applications.
- Specialized Examination Units: Establishing dedicated units with AI expertise would ensure consistent examination.
- Updated Guidelines: The Patent Office Manual should include specific guidance on examining AI-generated inventions, assessing inventive contribution and disclosure adequacy.[24]
International Cooperation
India should participate in international forums:
- WIPO Engagement: India should contribute to WIPO’s AI and IP policy development, advocating positions reflecting developing country interests.
- Bilateral Cooperation: India should engage with major jurisdictions through bilateral IP dialogues to explore harmonization possibilities.
Public Interest Safeguards
Any framework must incorporate public interest safeguards:
- Exclusions for Essential Technologies: AI-generated inventions in fields critical for public health should be subject to heightened scrutiny and broader exceptions.[25]
- Transparency Requirements: Public registries should clearly identify patents involving AI-generated inventions, enabling tracking of AI’s impact on patenting patterns.
- Periodic Review Mechanisms: Legislation should include mandatory review provisions requiring parliamentary reassessment within specified timeframes.
conclusion
The legal status of AI-generated inventions under Indian patent law remains fundamentally uncertain. This research demonstrates that current Indian patent law does not accommodate AI systems as inventors. Statutory provisions defining “inventor,” procedural requirements, and the broader legal architecture all presuppose human agency.
The Indian Patent Office’s rejection of DABUS applications represents defensible statutory interpretation consistent with the Patents Act. Absent explicit legislative authorization, recognizing AI as inventors would constitute impermissible judicial law-making. Comparative analysis reinforces this conclusion, as major patent systems worldwide have concluded that fundamental changes should originate from legislative action.
However, this research reveals significant policy concerns with categorically denying protection to AI-generated inventions. Patent law’s core objective—promoting innovation—may be frustrated if valuable AI-generated innovations cannot receive protection. The resulting immediate entry into public domain may reduce investment incentives in AI research.
The tension between statutory interpretation and policy objectives necessitates legislative intervention. Parliament should comprehensively address AI-generated inventions through amendments or creation of sui generis protection regimes. Such legislation should balance incentivizing AI innovation, ensuring disclosure, maintaining accountability, and protecting public interest, particularly regarding access to essential technologies.
Any legal reform must be informed by India’s developmental priorities. India’s patent regime has historically emphasized balancing innovation incentives with access to essential technologies. This balance must be preserved when addressing AI-generated inventions. Legislative reforms should prevent AI-generated patent portfolios from creating barriers to technology access or exacerbating inequalities.
The international landscape’s fragmentation creates challenges but also provides India opportunities to develop distinctive approaches reflecting its priorities. India need not simply adopt frameworks from developed jurisdictions but can pioneer approaches that advance AI innovation while robustly protecting public interests, potentially providing models for other developing countries.
Rupal Barjatya
Symbiosis Law School, Pune
[1] The Patents Act, § 2(1)(y), § 6, No. 39 of 1970, INDIA CODE (1970).
[2] Ryan Abbott, I Think, Therefore I Invent: Creative Computers and the Future of Patent Law, 57 B.C. L. REV. 1079, 1080-85 (2016).
[3] Thaler v. Comptroller-General of Patents, Designs and Trade Marks [2021] EWCA Civ 1374 (Eng.); Thaler v. Vidal, 43 F.4th 1207 (Fed. Cir. 2022); EPO Decision J 8/20 (Dec. 21, 2021).
[4] The Patents Act, § 3, No. 39 of 1970, INDIA CODE (1970).
[5] Novartis AG v. Union of India, (2013) 6 SCC 1 (India).
[6] Rebecca S. Eisenberg, Patents and the Progress of Science: Exclusive Rights and Experimental Use, 56 U. CHI. L. REV. 1017, 1024-28 (1989).
[7] Ryan Abbott, The Artificial Inventor Project, WIPO MAGAZINE, Dec. 2019, at 2-7.
[8] Thaler v. Comptroller-General of Patents, Designs and Trade Marks [2023] UKSC 49 (Eng.).
[9] Thaler v. Commissioner of Patents [2021] FCA 879 (Austl.).
[10] [13] Thaler v. Commissioner of Patents [2022] FCAFC 62 (Austl.); Thaler v. Commissioner of Patents [2022] HCATrans 191 (Austl.).
[11] EPO Legal Bd. App., Decision J 8/20 (Dec. 21, 2021).
[12] Prashant Reddy, Can AI Be an Inventor Under Indian Patent Law?, SPICY IP (Aug. 15, 2021), https://spicyip.com/2021/08/can-ai-be-an-inventor-under-indian-patent-law.html.
[13] The Patents Act, § 2(1)(s), No. 39 of 1970, INDIA CODE (1970).
[14] The General Clauses Act, § 3(42), No. 10 of 1897, INDIA CODE (1897).
[15] The Patents Act, § 2(1)(y), No. 39 of 1970, INDIA CODE (1970).
[16] The Patents Rules, Rule 13, No. 2003, INDIA CODE (2003).
[17] Controller Gen. of Patents, Designs and Trade Marks, Order in Application No. 202017019068 (Apr. 2021).
[18] The Patents Act, § 120, No. 39 of 1970, INDIA CODE (1970).
[19] Thaler, [2023] UKSC 49.
[20] EPO Decision J 8/20.
[21] Robert C. Scheinfeld & Jamie L. Spiegel, Intellectual Property Issues in Artificial Intelligence, 21 LANDSLIDE 30, 32-33 (2019).
[22] The Patents Act, § 2(1)(s), No. 39 of 1970, INDIA CODE (1970).
[23] The Patents Act, §§ 2(1)(j), 2(1)(ja), No. 39 of 1970, INDIA CODE (1970).
[24] INDIAN PATENT OFF., MANUAL OF PATENT OFFICE PRACTICE AND PROCEDURE ch. 8 (2019).
[25] The Patents Act, § 84, No. 39 of 1970, INDIA CODE (1970).