Introduction
The question of whether an artificial intelligence system can be recognised as the inventor of a patented invention is no longer a theoretical curiosity confined to academic seminars. The DABUS litigation, originating from Stephen Thaler’s applications filed across multiple patent jurisdictions listing his AI system as the sole inventor, forced patent offices and courts worldwide to confront a lacuna that legislators never anticipated when they drafted foundational patent statutes. The outcomes were almost universally unfavourable to Thaler, yet the legal reasoning behind those outcomes diverges considerably across jurisdictions, and the divergence is instructive precisely because it reveals that courts and offices are not yet operating from a coherent global framework for AI-generated innovation.
For India, the implications of the DABUS controversy are particularly consequential. India’s Patents Act 1970, drafted in a context where the sole question of inventorship concerned human beings acting individually or collectively, contains definitional provisions that will need to be examined afresh as AI systems become embedded in research workflows across pharmaceuticals, biotechnology, software, and engineering. The Controller General of Patents, Designs and Trade Marks has not yet formally addressed AI inventorship in any published examination guideline, leaving an interpretive vacuum that applicants, practitioners, and policy makers must now urgently fill.
This article examines the DABUS decisions across jurisdictions, maps the Indian statutory framework onto those decisions, analyses the policy arguments for and against recognising AI as a legal inventor, and proposes legislative reforms that would allow India to attract AI-driven innovation without abandoning the human-centred accountability that the patent system ultimately requires.
Legal Framework
The DABUS Applications: A Summary of Jurisdictional Outcomes
Stephen Thaler’s AI system, DABUS (Device for the Autonomous Bootstrapping of Unified Sentience), was identified as the named inventor in patent applications relating to a food container with a fractal surface and a flashing light device for attracting attention. Thaler’s position was consistent: he had not conceived the inventions; DABUS had; and intellectual honesty required naming the actual inventor.
The United Kingdom Intellectual Property Office refused the application on the ground that Section 7 of the Patents Act 1977 requires an inventor to be a person, a requirement upheld by the UK Supreme Court in Thaler v. Comptroller-General of Patents (2023). The Court’s majority held that the statutory language was unambiguous and that it was for Parliament, not the courts, to expand the definition of inventor. The European Patent Office similarly refused, holding that Article 81 of the European Patent Convention requires that an inventor be a natural person, and that an AI system could be neither an inventor nor have successor-in-title for the purpose of Article 60. The United States Patent and Trademark Office refused registration, a position subsequently affirmed by the Fourth Circuit in Thaler v. Vidal (2022), which held that the word “individual” in the Patent Act means a natural person under established statutory interpretation canons reinforced by the Dictionary Act. The Australian Federal Court initially allowed the application, reasoning that the Patents Act 1990 did not expressly restrict inventorship to humans, but the Full Federal Court reversed that decision in 2022, aligning Australia with the majority position. South Africa’s Companies and Intellectual Property Commission (CIPC), operating under a relatively formalistic examination system, granted a patent naming DABUS as inventor in 2021, making it the first such grant worldwide, though South African law does not subject applications to substantive examination, significantly limiting the precedential weight of that grant.
India’s Statutory Position
The Patents Act 1970, as amended by the Patents (Amendment) Act 2005, defines “inventor” in Section 2(1)(y) as “a person who is the actual deviser of the invention.” The word “person” in Indian law, absent specific statutory extension, ordinarily refers to a natural person or a legal person such as a company or registered entity. An AI system possesses neither natural personhood nor any form of legal personhood under Indian law. The Information Technology Act 2000 creates no category of legal personhood for AI systems. Accordingly, on a plain reading of Section 2(1)(y), DABUS and systems like it could not be named as inventors under Indian law without legislative amendment.
Section 3(k) of the Patents Act 1970 further complicates matters by excluding from patentability “a mathematical or business method or a computer programme per se or algorithms.” The CGPDTM’s Guidelines for Examination of Computer Related Inventions, revised in 2017, attempt to distinguish between a computer programme per se (not patentable) and an invention that uses a computer programme as a component to achieve a technical effect (potentially patentable). AI-generated inventions whose technical character is substantive, rather than residing entirely in the software, might survive Section 3(k) scrutiny if the human applicant frames the claim around the technical output rather than the algorithmic process. However, the inventorship problem under Section 2(1)(y) remains distinct from and independent of the patentability question under Section 3(k).
The European Patent Convention Model
Article 60 of the European Patent Convention vests the right to a European patent in the inventor and their successors in title. The EPO Boards of Appeal have consistently interpreted “inventor” to mean a natural person. The EPO’s reasoning in the DABUS decisions was that only a natural person can be an inventor because the concept of succession-in-title implies an original right that must vest somewhere, and only a natural human being can be the progenitor of that right. The structural elegance of this reasoning is that it avoids closing the door on AI involvement in invention: an AI system can contribute to an inventive process, but the natural person who directs, designs, trains, or operates the AI system remains the inventor if a sufficiently meaningful intellectual contribution to the conception of the invention can be attributed to that person.
Judicial Developments
The DABUS litigation produced no Indian court decisions because no equivalent application was filed before the Indian Patent Office. However, Indian courts have addressed related questions that illuminate how they might approach AI inventorship. The Madras High Court’s decisions on computer-related inventions, including the Ferid Allani v. Union of India line of cases (affirmed by the Delhi HC in 2019), established that the prohibition in Section 3(k) does not apply if the invention has a technical character beyond the software itself. This interpretive approach, borrowed in part from the EPO’s technical effect doctrine, is relevant because an AI-generated invention with a clear technical character might survive patent office scrutiny if the human-centred inventorship requirement is satisfied through a theory of the AI as tool rather than AI as inventor.
The Delhi High Court’s broader intellectual property jurisprudence reflects receptivity to technologically sophisticated arguments. In cases involving software patents and database rights, the Court has demonstrated willingness to adapt statutory interpretation to technological realities. It is therefore plausible that, if confronted with an AI-assisted invention where the AI contributed significantly but a human directed the research agenda and evaluated the output, an Indian court would uphold a patent naming the human as inventor rather than insisting on a rigid conception of unaided human creativity.
Contemporary Issues and Analysis
The Innovation Incentive Argument
The traditional justification for the patent system is that the prospect of a limited monopoly incentivises inventors to invest in the costly and risky process of innovation and to disclose their inventions to the public in return for that monopoly. Applied to AI, the argument faces an obvious difficulty: an AI system has no subjective states, no preferences about future outcomes, and no capacity to be incentivised. It cannot be rewarded in any meaningful sense. Accordingly, the innovation incentive rationale, as traditionally formulated, provides no support for vesting patent rights in AI systems directly.
What the incentive argument does support, properly reformulated, is the proposition that the owners and operators of AI systems should have adequate legal certainty about the ownership of AI-generated inventions, because that certainty will encourage investment in the deployment of AI for inventive purposes. If the output of an AI system cannot be patented at all because no human conceived the invention in the relevant sense, the consequence may be that AI-generated innovations are protected instead as trade secrets, withdrawn from public disclosure, and the public thus loses the benefit of the patent bargain.
The Corporate Ownership Argument
A structurally preferable approach, and one consistent with existing law in most jurisdictions, is to treat AI systems as sophisticated instruments and to vest patent rights in the human or corporate entity that deploys the AI for inventive purposes, in the same way that inventions made by employees in the course of their employment vest in the employer under Section 6(1)(a) of the Patents Act 1970. This corporate ownership model avoids the need to recognise AI as a legal person while still ensuring that AI-generated innovations are patentable and that incentives for investment in AI research tools remain intact.
The challenge with this approach is determining what quantum of human contribution is required for the human deployer to be named as inventor. If the human merely switches on the AI and retrieves the output, it is difficult to argue that the human has devised the invention in any meaningful sense. Conversely, if the human designs the research problem, curates the training data, sets the evaluation criteria, and selects the commercially viable output from among the AI’s proposals, the human’s intellectual contribution may be sufficient to sustain inventorship in the traditional sense.
The AI as Tool Doctrine
Several commentators and the USPTO’s February 2024 guidance on AI-assisted inventions have articulated what might be called the “AI as tool” doctrine: an AI system is to be treated as a sophisticated instrument, analogous to a microscope or a computer, and the human who uses the AI to arrive at an invention is the inventor. The critical question under this doctrine is whether the human has made a “significant contribution” to the conception of the invention, not merely to the identification of a problem or the deployment of a tool to generate solutions.
India’s CGPDTM has issued no guidance on AI-assisted inventions as of 2025. This silence creates uncertainty for applicants who use AI tools in their research and development processes, which is an increasingly large proportion of patent applicants in the pharmaceutical, biotechnology, and software sectors.
Comparative and International Perspective
The United States Patent and Trademark Office’s February 2024 guidance, “Inventorship Guidance for AI-Assisted Inventions,” represents the most detailed regulatory position adopted by any major patent office. The USPTO’s position is that AI cannot be an inventor but that inventions created with the assistance of AI are patentable provided a natural person has made a “significant contribution to the conception” of each claim. The guidance identifies three non-exhaustive factors derived from Pannu v. Iolab Corp. (1998): the person must contribute to the conception of the invention, the contribution must not be the result of merely following the instructions of another, and the person must appreciate what they have invented.
The World Intellectual Property Organization’s Conversation on Intellectual Property and Artificial Intelligence, which has been ongoing since 2019, has not produced a binding treaty, but the Issues Papers published by WIPO reveal that member states are divided between those who favour creating a new category of AI-generated invention with reduced protection (shorter term, lower inventive step requirement) and those who prefer adapting existing human-centred frameworks through liberal interpretation of the “significant contribution” standard.
China’s National Intellectual Property Administration issued revised guidelines in 2023 indicating that AI systems cannot be named as inventors but that inventions substantially assisted by AI are patentable when a natural person is named as inventor and the human’s contribution to conception is documented. China’s approach is functionally similar to the USPTO’s guidance, reflecting convergence among the largest patent-granting economies around the “AI as tool” model.
Practical and Policy Implications
For Indian patent practitioners, the immediate implication of the current legal framework is straightforward: applications must name a natural person as inventor, regardless of the actual degree of AI contribution to the inventive process. Practitioners should advise clients to document human involvement in the inventive process contemporaneously, recording decisions about the scope of the research problem, the selection of training datasets, the evaluation of AI-generated candidate inventions, and the translation of AI outputs into patent claims. This documentation serves both compliance and litigation functions.
For the CGPDTM, the policy implication is that examination guidelines urgently require updating to address AI-assisted inventions. Without guidance, examination outcomes will be inconsistent and unpredictable, deterring applicants who use AI tools from seeking Indian patent protection.
For the pharmaceutical and biotechnology sectors, which are the largest users of AI in inventive processes in India, clarity on inventorship requirements is particularly important because patent protection in these sectors is economically essential to recouping research investment. Ambiguity about the validity of patents for AI-assisted drug discovery creates downstream risks for licensing, investment, and regulatory approval processes.
Suggestions and Reforms
India should amend the Patents Act 1970 to introduce a clear statutory framework for AI-assisted inventions. The amendment should proceed on three principles.
First, the definition of “inventor” in Section 2(1)(y) should be amended to specify that an inventor must be a natural person but that inventions conceived with the substantial assistance of an AI system are patentable provided the named human inventor has made a significant intellectual contribution to the conception of the claimed invention. The amendment should not define “significant contribution” exhaustively but should authorise the CGPDTM to issue guidelines specifying relevant factors, including the design of the research problem, the curation of input data, and the evaluation and selection of AI-generated outputs.
Second, the Patents Act should introduce a mandatory disclosure requirement for AI-assisted inventions, requiring applicants to disclose the use of AI systems in the inventive process and to identify the human decisions that constituted the human inventor’s significant contribution. This disclosure requirement serves both accountability and prior art purposes, enabling examiners to assess inventive step claims in light of the AI system’s documented capabilities.
Third, Section 3(k) should be amended to clarify its interaction with AI-assisted inventions. The current exclusion of computer programmes per se should be retained, but a statutory carve-out should confirm that an invention conceived through the use of an AI system is not excluded from patentability under Section 3(k) solely because the inventive process involved an AI algorithm, provided the claimed invention has a technical character beyond the AI process itself.
Conclusion
The DABUS decisions have established, with near-universal consensus among the world’s major patent jurisdictions, that AI systems cannot be inventors under existing statutory frameworks. The decisions leave open, however, the more commercially significant question of how patent law should treat inventions in which AI systems play a substantial creative role alongside directing human beings. India’s Patents Act 1970 is silent on this question, and the CGPDTM has yet to address it in examination guidelines.
The stakes are high. India’s pharmaceutical, biotechnology, and technology sectors are increasingly deploying AI in research and development, and the legal certainty of patent protection for AI-assisted inventions is becoming a material factor in investment decisions. India can attract AI-driven innovation and protect its globally significant generic medicines industry only if its patent law provides a framework that is clear, consistent with international norms, and adapted to the realities of AI-assisted research. The amendments proposed in this article offer a path toward that framework, anchored in human accountability, institutionally practicable, and aligned with the direction that the world’s leading patent offices and courts are already moving.
