Authored By: Revati Hukke
NC Law College, Nanded, Maharashtra
I. INTRODUCTION
Abstract
The rapid advancement of artificial intelligence has disrupted the foundational assumptions upon which patent law across the world was built. From the patent offices of New Delhi to Washington D.C. and London, a single unsettling question looms: can a machine be an inventor? As AI systems independently generate technical innovations, the legal frameworks of India, the United States, and the United Kingdom have been found critically wanting. This article examines the gaps in existing patent law that render AI-generated inventions legally ambiguous, and analyses the downstream consequences of this ambiguity on startups and technology businesses that rely on such innovations for their competitive survival. Through a comparative analysis of statutory provisions — India’s Patents Act, 1970; the United States Patent Act (35 U.S.C.); and the United Kingdom’s Patents Act, 1977 — this article argues that the failure to recognise AI as an inventive entity, or to provide a clear alternative framework, creates a dangerous legal vacuum that stifles innovation, discourages investment, and leaves emerging businesses dangerously exposed.
In 2020, a landmark legal battle began when Dr. Stephen Thaler filed patent applications across multiple jurisdictions listing ‘DABUS’ — an AI system — as the sole inventor of two innovations it had autonomously generated. Patent offices in India, the United States, and the United Kingdom all rejected these applications, not because the inventions lacked novelty or utility, but for a far more fundamental reason: the law, as written, does not recognise a machine as an inventor.
This episode exposed a fault line in global intellectual property law that has only grown wider with the accelerating capabilities of generative AI, large language models, and autonomous systems. Today, AI tools are not merely assisting human inventors — they are increasingly generating technical solutions, drug compounds, software architectures, and engineering designs with minimal human creative input. The legal infrastructure that governs who may claim, own, and profit from such inventions has simply not kept pace.
For established corporations with dedicated legal teams and lobbying power, navigating this ambiguity is a costly inconvenience. For startups and small technology businesses — who often depend entirely on a single AI-derived innovation as their core value proposition — it can be existential. Without clear patent protection, a startup cannot raise venture capital, cannot defend against infringement, and cannot prevent a larger competitor from freely copying its core technology.
This article proceeds in five parts. Part II sets out the existing legal frameworks in India, the United States, and the United Kingdom. Part III identifies the critical gaps in each jurisdiction. Part IV analyses the specific impact of these gaps on startups and emerging technology businesses. Part V offers reform proposals and comparative reflections. The article concludes with a call for urgent legislative intervention.
II. EXISTING LEGAL FRAMEWORKS: A COMPARATIVE OVERVIEW
A. India: The Patents Act, 1970
India’s primary legislation governing patents is the Patents Act, 1970, as amended by the Patents (Amendment) Act, 2005. Section 2(1)(y) defines an ‘inventor’ as ‘a person who is the actual deviser of the invention.’ The use of the word ‘person’ is central to the current controversy. Under Indian law, and consistent with the general principles of legal personality under the Indian Contract Act, 1872 and related statutes, an AI system is not a legal person and therefore cannot qualify as an inventor.
The Indian Patents Act further requires, under Section 6, that only the ‘true and first inventor’ or their assignee may apply for a patent. Where an AI system has generated an invention without meaningful human creative input, it becomes impossible to identify a human inventor in good faith. The Indian Patent Office has issued no specific guidelines on AI-generated inventions, and Indian courts have yet to directly confront this question, leaving the matter in a state of uncomfortable legal silence.
It is worth noting that India’s patent law already contains several restrictions on patentability — most notably, Section 3(k), which excludes computer programmes per se and mathematical methods from patent protection. While this provision was primarily directed at software patents, its uncertain interaction with AI-generated algorithmic inventions adds another layer of complexity for technology businesses seeking protection in India.
B. The United States: The Patent Act (35 U.S.C.)
The United States Patent Act defines an inventor under 35 U.S.C. § 100(f) as ‘the individual or, if a joint invention, the individuals collectively who invented or discovered the subject matter of the invention.’ The use of the term ‘individual’ has been interpreted by the United States Patent and Trademark Office (USPTO) and the federal courts as requiring a natural person.
This interpretation was squarely confirmed in Thaler v. Vidal, 43 F.4th 1207 (Fed. Cir. 2022), where the United States Court of Appeals for the Federal Circuit held that ‘inventors must be natural persons’ under existing law and that DABUS could not be listed as an inventor. The court explicitly noted that the issue of whether AI should be granted inventorship was a matter for Congress, not the courts, to resolve.
The USPTO has since issued guidance affirming that AI-assisted inventions may still be patentable, provided that a natural person made a ‘significant contribution’ to the conception of the invention. However, this standard — borrowed from joint inventorship doctrine — is notoriously difficult to apply in the context of AI, where the human’s contribution may be limited to selecting training data, defining a general problem, or choosing among AI-generated outputs.
C. The United Kingdom: The Patents Act, 1977
The United Kingdom’s Patents Act, 1977 defines an inventor under Section 7(3) as ‘the actual deviser of the invention.’ In Thaler v. Comptroller-General of Patents [2023] UKSC 49, the United Kingdom Supreme Court unanimously held that DABUS could not be listed as an inventor under the 1977 Act. The Court found that the Act unambiguously required a human inventor, and that its role was to interpret the existing statute, not to extend it.
What distinguishes the UK position is that the Supreme Court explicitly acknowledged the policy vacuum created by this interpretation and suggested that Parliament should consider the matter. The UK Intellectual Property Office (UKIPO) has since launched consultations on AI and IP, and in its 2022 consultation response, it noted that the current framework may be inadequate for addressing AI-generated works — though no legislative reform has yet followed.
The UK also benefits from a provision under Section 39 of the Patents Act, 1977 that addresses employee inventions, which has been suggested as a potential analogy for assigning ownership of AI-generated inventions to the deployer of the AI system. However, this remains an untested proposition, and its application to autonomous AI-generated innovations is far from settled.
III. THE CRITICAL GAPS: WHERE THE LAW FAILS
A. The Inventorship Vacuum
The most fundamental gap across all three jurisdictions is the absence of a legal category for AI-generated inventions. The law offers two choices: a human inventor, or no patent. This binary is increasingly divorced from the reality of modern innovation. When an AI system generates a patentable invention with no more human involvement than the setup of a computational environment, there is no honest human inventor to name — yet the law requires one.
This creates a perverse incentive for applicants to inflate the human contribution to an AI-assisted invention, listing a nominal human inventor even where the substantive creative work was performed by the AI system. Such practices undermine the integrity of the patent system and create legal risk for businesses that may later find their patents invalidated on grounds of incorrect inventorship.
B. The Ownership Ambiguity
Even where inventorship can be notionally attributed to a human, the question of who owns the resulting patent remains unresolved. In each jurisdiction, patent ownership follows from inventorship: the inventor owns the patent, subject to assignment. If AI systems are excluded from inventorship, it is unclear whether the person who designed the AI, the person who deployed it, the person who defined the problem it was asked to solve, or some combination of these parties should be considered the owner.
This ambiguity is particularly damaging for startups, which may have contracted with an external AI service provider, licensed an AI platform from a major technology company, or built upon open-source AI tools. In each of these scenarios, the startup’s claim to own the resulting invention is legally uncertain, making it difficult to secure patent protection, attract investment, or defend against infringement.
C. The Disclosure Problem
Patent law in each jurisdiction requires that a patent application disclose the invention in sufficient detail to enable a person skilled in the relevant field to reproduce it. In the context of AI-generated inventions, this requirement creates significant practical difficulties. The ‘invention’ may emerge from a complex, non-linear AI process that cannot be meaningfully reduced to a human-legible technical description. The underlying AI model may itself be proprietary, making full disclosure impossible without revealing trade secrets.
Neither the Patents Act, 1970, the US Patent Act, nor the UK Patents Act, 1977 contains any provision adapted to the disclosure challenges posed by AI-generated inventions. Startups that rely on AI-generated innovations are thus faced with an impossible choice: disclose enough to meet the enablement requirement and potentially expose their proprietary AI systems, or withhold information and risk invalidation of the patent.
IV. IMPACT ON STARTUPS AND TECHNOLOGY BUSINESSES
A. The Investment and Valuation Crisis
Patent portfolios serve a critical function in the startup ecosystem. For early-stage technology companies, patents are not merely legal protections — they are signals of innovation, barriers to competition, and often the primary assets that justify venture capital investment. When a startup’s core innovation is AI-generated, and patent protection for that innovation is legally uncertain, its entire valuation model may be undermined.
Investors performing due diligence on a startup’s IP portfolio increasingly encounter the question of whether AI-generated innovations are properly protected. In the absence of clear legal guidance, many investors apply a discount to AI-derived patents, or require startups to restructure their IP strategies to emphasise human contributions — even where such contributions are marginal. This creates a structural disadvantage for AI-driven startups relative to competitors in jurisdictions that may develop clearer AI patent frameworks.
B. Cross-Border Vulnerability
Many technology startups operate across multiple jurisdictions simultaneously, seeking patent protection in India, the United States, and the United Kingdom as part of a global IP strategy. The divergent and inconsistent treatment of AI-generated inventions across these three jurisdictions creates a compliance burden and a litigation risk that disproportionately affects smaller businesses.
A startup that succeeds in obtaining a patent in one jurisdiction — perhaps by emphasising human contributions that satisfy that jurisdiction’s inventorship standard — may find its patent challenged or invalidated in another jurisdiction applying a different standard. The lack of international harmonisation compounds this vulnerability, given the absence of any binding guidance from the World Intellectual Property Organization (WIPO) specifically addressing AI inventorship.
C. Competitive Inequality
Large technology corporations that deploy AI extensively — Google, IBM, Microsoft, and their peers — possess in-house legal teams capable of navigating inventorship ambiguity, restructuring patent applications, and pursuing alternative protections such as trade secrets. Startups, by contrast, typically lack these resources. The legal vacuum created by outdated patent frameworks therefore operates asymmetrically, protecting large incumbents while leaving smaller innovators exposed.
This dynamic risks entrenching the very technology monopolies that competition law in each jurisdiction purports to prevent. If only large corporations can effectively protect AI-generated innovations, they will accumulate vast patent portfolios that create insurmountable barriers to entry for startups — precisely the opposite of the incentive structure that patent law was designed to create.
V. REFORM PROPOSALS AND COMPARATIVE REFLECTIONS
The inadequacy of existing frameworks demands legislative intervention. Several reform models merit consideration:
- A dedicated ‘AI-generated invention’ category could be introduced in each jurisdiction’s patent law, permitting a natural person — the deployer, commissioner, or operator of the AI system — to be recognised as the legal inventor of an AI-generated innovation, subject to disclosure requirements adapted to AI processes.
- Drawing on Section 39 of the UK Patents Act, 1977, each jurisdiction could adopt a ‘controller model’ of AI inventorship, in which patent rights vest automatically in the person who directed or deployed the AI system to solve the problem that generated the invention. This approach would resolve the ownership ambiguity without requiring a fundamental reconception of inventorship doctrine.
- International harmonisation through WIPO is essential. The current divergence between jurisdictions creates opportunities for forum shopping and regulatory arbitrage that benefit large corporations at the expense of smaller innovators. A WIPO treaty or model law on AI-generated inventions would provide the consistency that a globalised innovation economy requires.
India, in particular, has an opportunity to position itself as a progressive jurisdiction in this space. With a growing startup ecosystem and a government committed to technology-led development, India could attract AI-driven innovation by adopting a forward-looking AI patent framework ahead of its competitors. The absence of any existing judicial precedent directly on point means that legislative reform faces fewer institutional obstacles in India than in the United States or the United Kingdom — making the comparative case for early action all the more compelling.
VI. CONCLUSION
Patent law was built for a world in which only human minds could invent. That world no longer exists. Across India, the United States, and the United Kingdom, the statutory frameworks that govern patent protection have been confronted with AI-generated innovations and found inadequate. The consequences of this inadequacy fall most heavily on the startups and technology businesses that are at the forefront of the AI revolution — the very innovators that patent law was intended to nurture.
The DABUS litigation was not merely a curiosity. It was a warning. As AI systems grow more capable and more autonomous, the gap between what the law contemplates and what innovation requires will widen further. Legislative reform — whether through dedicated AI inventorship provisions, controller-based ownership models, or international harmonisation — is not merely desirable. It is urgent.
The question is no longer whether the law must change. The question is whether lawmakers will act before the damage to the startup ecosystem, and to the integrity of the patent system itself, becomes irreversible.
REFERENCE(S):
Legislation
Patents Act 1970 (India), ss 2(1)(y), 3(k), 6.
Patents Act 1977 (UK), ss 7(3), 39.
United States Patent Act, 35 U.S.C. §§ 100(f), 101, 102, 103, 112.
Cases
Thaler v Vidal, 43 F.4th 1207 (Fed. Cir. 2022).
Thaler v Comptroller-General of Patents, Designs and Trade Marks [2023] UKSC 49.
Thaler v Commissioner of Patents [2021] FCA 879 (Federal Court of Australia, discussed for comparative context).
Secondary Sources
Ryan Abbott, The Reasonable Robot: Artificial Intelligence and the Law (Cambridge University Press 2020).
Andres Guadamuz, ‘Artificial Intelligence and Copyright’ (2017) WIPO Magazine, October 2017.
UK Intellectual Property Office, ‘Artificial Intelligence and Intellectual Property: Copyright and Patents — Response to Consultation’ (June 2022).
USPTO, ‘Inventorship Guidance for AI-Assisted Inventions’ (February 2024) 89 Fed. Reg. 10043.
WIPO, ‘WIPO Conversation on Intellectual Property (IP) and Artificial Intelligence (AI)’ (WIPO/IP/AI/3/GE/20) (2020).
Adarsh Ramanujan and Aditya Gupta, ‘Patenting Artificial Intelligence in India: Issues and Challenges’ (2021) 26(2) Journal of Intellectual Property Rights 85.
Russ Pearlman, ‘Recognizing Artificial Intelligence (AI) as Authors and Inventors Under U.S. Intellectual Property Law’ (2018) 24 Richmond Journal of Law and Technology 2.
