This article synthesizes research conducted on the basis of Professor Na Jong-gap's (Yonsei University) work on "Philosophical Approaches to AI Inventions." As the age of artificial intelligence accelerates, the question of "Can AI be recognized as an inventor?" is no longer confined to philosophical speculation. Courts and patent offices in the United States, Europe, and Korea are now confronting this issue directly. The deepest answer to that question may be found not in statutory text, but in Immanuel Kant's philosophy of property rights and the utilitarian and law-and-economics theories of incentive structures.

I. Three Kantian Grounds for Denying AI Inventorship

1. Absence of Autonomy and Personhood — Property Rights as a Right of Personality

Kant's concept of ownership and property rights originates from the personality (Persönlichkeit) of the autonomous human being — that is, from human autonomy itself. In Kantian philosophy, property rights serve as the ultimate means of securing the innate human right to freedom (Freiheit). Only a 'natural person' who holds dominion over their own personality and can express the will to possess may be recognized as a legal subject capable of holding rights.

Artificial intelligence possesses no autonomy of its own; it operates solely under human direction and control. It therefore cannot attain the status of an 'autonomous person' as Kant defines the subject of rights.

2. The Impossibility of 'Conception' as a Mental Act

The 'conception' of an invention — the subjective, private mental act driven by human autonomy — is the origin of every patent and the very essence of inventiveness. It is the process by which the 'nature of mind' of a natural person is individualized and made one's own. From a Kantian standpoint, no matter how capable an AI system may be at executing tasks, a machine cannot perform such subjective mental activity or engage in autonomous conception, and therefore cannot be recognized as an inventor under the law. What AI performs is, at most, mechanical work (Mechanical work) conducted under human control.

3. Status as an End-in-Itself and the Irrelevance of 'Freedom'

In Kantian ethics, human beings must be treated as ends-in-themselves (Zweck an sich selbst) — beings of inherent dignity, never mere means. Artificial intelligence, by contrast, is an artifact created to serve human needs; it cannot become an end unto itself and cannot hold the moral rights or property rights that humans possess.

Crucially, whereas Kant justifies property rights as a means of securing ultimate human freedom, machines such as AI have no need for 'freedom' in any meaningful sense. There is therefore no philosophical justification or existential purpose that could support granting them exclusive proprietary rights.

II. The Denial of AI Inventorship and the Incentive Structure of the Patent System

The denial of AI inventorship transcends mere statutory interpretation. It is organically linked to a legal-philosophical and law-and-economic necessity to defend the ultimate purpose of the patent system and its economic incentive structure. This connection can be analyzed along three key dimensions.

1) The Redundancy of Utilitarian and Pragmatist Incentive Structures

The utilitarian justification for granting inventors strong exclusive rights rests on the proposition that the patent system must reward the agonizing mental labor (Mental Labor), enormous cost, and sacrifice entailed in inventive activity, thereby actively incentivizing and encouraging the development of new technology.

  • Absence of motivation: Artificial intelligence — a non-human actor — does nothing more than mechanically repeat operations dictated by designed algorithms and computational resources. It is not an agent that acquires creative motivation or alters its behavior based on whether legal or economic monopoly rights are guaranteed. Legal incentives are entirely unnecessary for AI.
  • Loss of the system's foundational purpose: If exclusive patent rights were indiscriminately extended to cover inventions autonomously generated by AI without meaningful human intellectual intervention, patent law as an incentive structure designed to encourage human technological contribution and sacrifice would lose its raison d'être.

2) A Law-and-Economics Perspective: Preventing the 'Tragedy of the Anticommons'

From a law-and-economics standpoint, the patent system must balance the provision of innovation incentives against the imperative not to unduly encroach upon the Public Domain — the common intellectual heritage of humanity.

  • Formation of a Patent Thicket: If the countless incremental modifications and combinations mechanically generated by computationally powerful AI systems were each granted proprietary status (patent rights) indiscriminately, rights would become so fragmented as to produce a dense 'Patent Thicket.'
  • Resource paralysis and net social loss: Subsequent human innovators would be forced to negotiate with countless rights holders and pay astronomical licensing fees even to employ the most trivial technical element — a classic 'Tragedy of the Anticommons.' This would severely retard overall technological progress and social welfare.
  • Denial of subjecthood as a filtering mechanism: Denying AI inventorship per se, and permitting privatization only of those outputs that embody genuine, non-obvious human intellectual labor filtered through rigorous novelty and inventive-step requirements, constitutes an essential legal-institutional defensive mechanism for protecting the intellectual commons and maximizing social welfare (Kaldor–Hicks improvement).

3) Coherence with Natural Law and Kantian Autonomy Doctrine

In Kant's philosophy of property rights, the establishment of exclusive rights (property rights) is justified solely as a means of guaranteeing and securing the 'freedom (Freiheit)' of the autonomous person — a natural person who can express the will to own and make decisions autonomously.

  • Beings for whom freedom is irrelevant: Machines such as AI — which operate purely under human control and lack autonomous will or personhood — possess no moral subjectivity or 'freedom' that warrants protection. There is therefore no moral or philosophical basis on which to recognize private property rights (patents) in them.
  • Control as a tool and the right of intellectual dominion: AI cannot perform the subjective mental act at the core of inventorship — 'conception' — and amounts to nothing more than a sophisticated creative instrument. Accordingly, only the natural human person who exercises instrumental control over AI, leads the processes of selection, exclusion, and testing, and thereby establishes 'Intellectual Domination' over the entire creative act, can be recognized as the sole rightful inventor under patent law.

III. The Enablement Requirement — A Filter Against the Privatization of Mass AI Output

The 'Enablement Requirement' of patent law operates as the most critical legal filter for blocking the indiscriminate privatization (patenting) of the countless technical ideas and data points mechanically generated by AI, and for realizing the patent system's foundational purpose of complete public disclosure and broad dissemination of technical knowledge (the Quid Pro Quo bargain).

1) Technical Filtration Through 'No Undue Experimentation'

The Enablement Requirement is not a mere formality of putting words on a page in a specification. It is a substantive contractual gateway that underpins the entire patent system.

  • Strict control over reproducibility: To satisfy this requirement, a person having ordinary skill in the art (PHOSITA) must be able to reproduce (make and use) the invention clearly and readily, using only the specification's disclosure, without undue additional experimentation or research.
  • Blocking incomplete AI outputs: AI can mechanically generate tens of thousands of candidate compounds or raw data fragments at extraordinary speed, but these outputs typically constitute 'incomplete knowledge' — lacking precise causal relationships or concrete execution pathways. AI's raw outputs, which can only be completed after extensive additional human experimentation and research by a PHOSITA, inherently fail to satisfy the Enablement Requirement and are filtered out at the patent gateway.

2) A Subjectivity Filter Compelling Human 'Substantial Contribution' and 'Mental Labor'

As AI is increasingly used as a creative instrument, the Enablement Requirement becomes the decisive measure that compels substantive human intervention by a natural person.

  • The human as the organizer of technical knowledge: The sophisticated task of transforming AI-generated provisional data or initial ideas into a refined, complete technical form that a third party can readily reproduce — and of clearly disclosing it in the specification — is possible only through the mental labor (Mental Labor) of a natural human person.
  • The watershed of inventorship recognition: Only when a human takes the initiative to organize technical information derived from AI's mechanical outputs and to reduce it to a form that enables practice does that person qualify as a rightful inventor who has "substantially contributed to the creative act of a technical idea." Conversely, simply transcribing AI's mechanical output into a patent specification without such effort will fail the Enablement Requirement and constitute grounds for invalidity.

3) Post-Grant Invalidation and Anti-Monopoly Mechanisms Against Fraudulent Privatization

The Enablement Requirement operates not only at the examination stage, but also as a powerful sanction capable of stripping private monopolies over deficient AI outputs after grant.

  • Invalidation of fraudulently obtained patents: Any attempt to fraudulently obtain a patent by presenting AI-generated hypothetical data or unverified figures as though they were successful experimental results, or by otherwise violating the enablement requirement, renders the patent subject to clear grounds of invalidity.
  • Inequitable Conduct and unenforceability: Where inequitable conduct — such as concealing material technical information or making false representations in violation of the Duty of Candor — is proven, not only the individual claims but the entire patent becomes unenforceable. Furthermore, wielding a fraudulently obtained patent to block entry by generic competitors constitutes patent misuse under competition law, attracting severe administrative and criminal sanctions.
  • Criminal sanctions for false inventorship — completing the enforcement architecture: The duty to accurately identify inventors (accurate recording of inventor identity) is not a mere procedural formality — it is a substantive line of defense against distortion of inventorship attribution in the age of AI. Where it is established in inter partes review or litigation that AI was falsely listed as an inventor, or that a natural person with no substantive contribution was nominally designated as inventor, such conduct constitutes grounds for patent invalidity, and may further amount to False Oath or fraudulent patent prosecution, warranting criminal prosecution. The enforcement architecture against fraudulent privatization is thereby completed as a triple-deterrent structure: invalidity → unenforceability → criminal sanction.

Conclusion: Quid Pro Quo — Safeguarding the Public Domain

The exceptional privilege of a private monopoly must be justified by the complete public disclosure of knowledge capable of being returned to society as a whole.

By operating this filter forcefully, the patent system can practically forestall the 'Tragedy of the Anticommons (Patent Thicket)' whereby the trivial variants mass-produced by AI encroach upon the intellectual commons — and ensure that, upon expiration of the patent term, that knowledge passes fully into the Public Domain, promoting humanity's collective welfare and competitive innovation (Kaldor–Hicks improvement).

IV. Protection of Inventions Made Using AI as a Creative Tool — The Doctrine of 'Intellectual Domination'

Professor Na's research presents a highly consistent and sophisticated legal-philosophical position. Its core thesis is that while AI's own inventive subjecthood is categorically denied, the inventorship and patent rights of the natural human person who exercises control over AI as a tool — and who performs substantive mental labor — are fully recognized.

1. 'Intellectual Domination' and the Conception of an Invention

The most critical stage for the formation of an invention under patent law is 'Conception' — the moment at which a definite and complete solution to a recognized problem is formed in the mind of the inventor.

  • Maintaining intellectual domination: In the course of completing an invention, an inventor may receive ideas, materials, and suggestions from other persons or from machines (AI). However, so long as the inventor continues to maintain 'Intellectual Domination' over the creative work — through the processes of successful testing, selecting, and rejecting — they do not lose their status as inventor.
  • Individualization and subjectivization: From a Kantian philosophical perspective, the conception of an invention is the process by which the inventor draws the common assets of humanity (such as natural laws) into their own subjective 'space and time' through private, subjective mental activity and achieves dominion over them. Where a human reviews AI-generated outputs, makes selections from them, and concretizes them under their own rational control, that person has achieved intellectual domination and subjectivization — and the result is properly recognized as a human invention.

2. Kant's 'Autonomy' and the Doctrine of 'Instrumental Control'

  • AI as a tool lacking autonomy: AI possesses no autonomy of decision-making; it is simply a being that performs mechanical work under human direction and control. AI therefore cannot become an 'end-in-itself' rather than a mere means, and cannot hold moral rights or property rights.
  • Application of the camera precedent: In the landmark copyright case Burrow-Giles Lithographic Co. v. Sarony, the court held that "the camera as a machine cannot be an author, but the human who exercised creative control over the camera as an instrument is the author." By analogy, AI is likewise nothing more than a sophisticated instrument operated pursuant to human autonomous will — and the human who leads the creative process by wielding it holds the status of inventor.

3. Locke's 'Mental Labor' and the Requirement of Substantial Contribution

John Locke's labor theory of value, and the common law doctrine that succeeds it, aim to protect the fruits of 'Labors of the mind' — the result of human pain and effort.

  • The proper objects of reward: In utilitarian and pragmatist philosophy, the exclusive right of a patent is granted to incentivize human sacrifice and technological contribution. Non-human actors such as AI have no need whatsoever for such legal or economic inducement.
  • Substantial Contribution: Korean Supreme Court precedent likewise holds that to qualify as an inventor, one must have "substantially contributed to the creative act of a technical idea." Only when a human goes beyond the mere mechanical use of AI — actively proposing, supplementing, and refining new ideas based on AI assistance, or otherwise adding the substantive mental labor and contribution of a natural person — does the result qualify as a legally cognizable invention.

V. Concrete Legal Standards for 'Intellectual Domination' and Methods for Assessing 'Substantial Contribution'

1) Concrete Legal Standards for Proving a Human's 'Intellectual Domination'

The central inquiry for determining whether an invention has been made and whether a person qualifies as its inventor is whether the inventor maintained a state of 'intellectual domination' throughout the creative process. The following specific legal standards appear in precedents such as the USPTO Patent Trial and Appeal Board (PTAB) decision in Morse v. Porter:

  • Leading the processes of successful testing, selection, and rejection: An inventor may receive suggestions, ideas, and materials from various sources and incorporate them into the invention. But intellectual domination is maintained only where the human directly leads and controls the final decisions of what to select (selecting), what to reject (rejecting), and how to test and validate the result (successful testing).
  • Retaining control even when a key solution is adopted: Even where a suggestion from another person or machine proves to be the 'key that unlocks his problem,' the inventor does not lose inventorship status if they maintained mental dominion and decision-making authority over the work throughout.
  • Achieving 'subjectivization' and 'notional possession' in space and time: The inventor must review and select from AI's mechanically generated raw outputs and concretize them under their own autonomous will and rational control — thereby establishing a state of intellectual domination (notional possession, possessio noumenon) — for the result to be recognized as a human invention.

2) Methods for Assessing 'Substantial Human Contribution' to AI-Generated Outputs

Because AI lacks the free will to make independent decisions and operates purely as a sophisticated 'instrument' under human control, AI-generated outputs can only be protected as patents when the substantive mental labor and contribution of a natural human person has been added. The following criteria — derived from Korean Supreme Court precedent and legal principles — govern the assessment of substantial contribution:

  • Exclusion of mere mechanical use: A human's act of taking AI-generated outputs and filing them for patent protection without independent control or substantive human oversight does not constitute 'substantial contribution.' Just as the human who operates a camera is the author, one must exercise instrumental control over AI and project human subjective intent into the manner of its operation.
  • Concrete intervention in the creative act of a technical idea (Korean Supreme Court standard): Courts assess whether a person has substantially contributed to the creative act by applying the following behavioral criteria:
    1. Did the person newly propose, add to, or supplement a specific conception based on the provisional results or data generated by AI?
    2. Did the person, through additional experimentation or research, concretize AI's nascent conception into an actual solution?
    3. Did the person independently provide specific means and methods for achieving the purpose and effects of the invention, or render specific advice and guidance?
  • Taking the lead in satisfying the Enablement Requirement: Building on AI-generated knowledge, the person must be able to systematize technical information and clearly disclose it in the specification (satisfying the Enablement Requirement) in a manner that allows a PHOSITA to reproduce the invention clearly and readily without undue additional experimentation — only then is the result recognized as a complete invention to which genuine mental labor has been contributed.

Closing Remarks

Under Professor Na Jong-gap's scholarly analysis, when a human completes an invention by using AI as a tool, the result constitutes — so long as the human maintains 'intellectual domination' and 'autonomous control' — a morally and legal-philosophically intact human invention. This is a legitimate mode of acquiring rights that aligns perfectly with natural law theories of property, modern positive law, and prevailing judicial attitudes.

Patent law in the age of AI ultimately converges on a single question: "Who, and to what extent, maintained intellectual domination over the creative process?" The philosophical and legal answer to that question determines the attribution of AI-related inventions and constitutes the central pillar sustaining the legitimacy of the patent system — as this research has argued.

References

  1. Na, Jong-gap. (2024). Immanuel Kant's Philosophy of Property Rights and the Legitimacy of Patent Rights: Including a Philosophical Examination of AI's Status as an Inventive Subject. Justice, (203), 105–143.
  2. Na, Jong-gap. (2023). A Study of Patents, Patent Rights, and Patent Law: The Development of Natural Rights and Utilitarian Instrumentalism, and the Formation of Western Capitalist Economic Ethics (Yumin Series 23). Hongjin Foundation for Legal Research; Gyeongin Publishing.
  3. Na, Jong-gap. (2021). Locke, Spencer, Nozick, Pareto, and Kaldor–Hicks: Combining Natural Rights Justification and Utilitarian Justification for Patent Rights. Intellectual Property Rights, 66, 1–39. https://doi.org/10.36669/ip.2021.66.1
  4. Na, Jong-gap. (2010). The Development and Prospects of Theories on the Legitimacy of Patent Rights. Comparative Private Law, 17(1), 561–607.
  5. Na, Jong-gap. (2010). The Development of the Inventive Step Concept in Patent Law: The Tragedy of the Anticommons and Efficiency. Intellectual Property Rights, 32, 41–86.
  6. Na, Jong-gap. (2005). A Study on the Nature of Patents. Intellectual Property Rights, 17, 31–72.

Related Video & Audio

[YouTube] AI & Patent Law Philosophy — Kant's Theory and AI Inventorship (Lecture Video)
[Audio] Why Patent Law Does Not Recognize AI as an Inventor—and the Coming Risks of Secrecy