AI-Generated Inventions and Patent Ownership: A Comparative Analysis of Indian and International Jurisprudence

Subject Reference: Patents and Geographical Indications (IPR Law & Policy)

The intersection of Artificial Intelligence (AI) and Intellectual Property (IP) Law is undergoing an unprecedented structural stress test. Historically, patent systems were engineered around an axiomatic premise: the inventor is an individual human being whose intellectual labour, ingenuity, and inventive leap justify a temporary statutory monopoly.

Today, advanced machine learning architectures, generative diffusion models, and autonomous discovery pipelines (such as AlphaFold, generative molecular design engines, and evolutionary neural algorithms) no longer function merely as passive calculating instruments. They actively formulate hypotheses, synthesize novel chemical entities, optimize antenna geometries, and design complex mechanical structures without direct human step-by-step guidance.

This technological evolution raises fundamental legal questions:

1. Can an AI system be legally recognized as an “inventor” on a patent application?
2. If an AI generates an invention autonomously, who owns the resulting intellectual property rights?
3. How does the widespread adoption of AI affect statutory thresholds like the Person Skilled in the Art (PSITA) and Sufficiency of Disclosure?
4. How do Indian statutory provisions compare with international legal precedents set by the USPTO, EPO, and the UK Supreme Court?

1. The Core Legal Conundrum: Inventorship vs. Ownership

To navigate the jurisprudence of AI patents, one must delineate the strict boundary between Inventorship and Ownership:

• Inventorship is a matter of factual and legal attribution. It recognizes the natural person(s) who conceived the essential, operative inventive concept embodied in the patent claims. Inventorship carries inalienable moral and statutory recognition.
• Ownership relates to proprietary and economic exploitation rights—the capacity to hold, license, assign, pledge, or enforce the patent.

Under classical jurisprudence, ownership flows from inventorship. An employer, corporate entity, or assignee cannot hold initial title to an invention out of thin air; their title derives from an assignment executed by the original human inventor who possessed legal capacity. Because AI systems lack legal personhood, they cannot hold property, contractually assign rights, or sue/be sued.

┌─────────────────────────────────────────────────────────────┐
│                   INVENTIVE PROCESS                         │
│                                                             │
│   [Human Ingenuity / AI Pipeline]                           │
│               │                                             │
│               ▼                                             │
│   INVENTORSHIP (Attribution)                                │
│   - Statutory recognition of conceiving inventive step      │
│   - Requires natural personhood across major jurisdictions  │
│               │                                             │
│               ▼ (Legal Assignment / Employment Contract)    │
│   OWNERSHIP (Proprietary Title)                             │
│   - Exploitation, licensing, enforcement                    │
│   - Vests in individuals, corporations, or assignees        │
└─────────────────────────────────────────────────────────────┘

2. The Indian Legal Framework: The Patents Act, 1970

India’s intellectual property regime is governed by the Patents Act, 1970 (as amended) and administered by the Office of the Controller General of Patents, Designs and Trade Marks (CGPDTM).

A. Who Can Apply for a Patent? (Sections 2 and 6)

Section 6 of the Patents Act, 1970 explicitly governs the eligibility to file a patent application:

Section 6(1): An application for a patent for an invention may be made by: (a) by any person claiming to be the true and first inventor of the invention; (b) by any person being the assignee of the person claiming to be the true and first inventor…; (c) by the legal representative of any deceased person who immediately before his death was entitled to make such an application.

The critical statutory phrases are “person” and “true and first inventor”:

• Section 2(1)(s) defines a “person” to include the government. Under general statutory interpretation in India (General Clauses Act, 1897), a person encompasses natural persons and legally recognized corporate entities (juristic persons). AI has not been granted juristic personhood under Indian civil law.
• Section 2(1)(y) clarifies that “true and first inventor” does not include either the first importer of an invention into India or a person to whom an invention is first communicated from outside India.

The legislative intent behind Section 6 and Section 2(1)(y) was to reward human mental ingenuity. An AI algorithm cannot swear the mandatory declaration of inventorship required under Form 1 (Application for Grant of Patent) or Form 5 (Declaration as to Inventorship).

B. Section 3(k) and Computer-Related Inventions (CRI)

Even if inventorship hurdles are resolved, Section 3(k) acts as a substantive bar for purely software- or algorithm-driven developments:

Section 3(k): The following are not inventions within the meaning of this Act: “a mathematical or business method or a computer programme per se or algorithms;”

Under the Revised Guidelines for Examination of Computer Related Inventions (CRI Guidelines, 2017) issued by the Indian Patent Office, patent examiners must ensure that:

1. The claim is not directed solely to an algorithm or mathematical formula.
2. The software or AI model produces a tangible technical effect or exhibits a technical contribution when combined with novel hardware.

If an AI generates an invention where the contribution lies merely in data modeling, statistical correlation, or an algorithmic discovery without tangible technical manifestation, it faces direct rejection under Section 3(k).

┌──────────────────────────────────────────────────────────────┐
│          INDIAN STATUTORY FILTERS FOR AI INVENTIONS          │
├──────────────────────────────────────────────────────────────┤
│ 1. SECTION 6 & 2(1)(y) : Must have a human "true and first   │
│                          inventor".                          │
│ 2. SECTION 3(k)        : Must not be an algorithm or software│
│                          "per se"; needs technical effect.   │
│ 3. SECTION 10(4)       : Must fully describe the best method │
│                          and disclose operative code/weights.│
│ 4. SECTION 2(1)(ja)    : Must demonstrate an Inventive Step  │
│                          not obvious to a PSITA.             │
└──────────────────────────────────────────────────────────────┘

3. Global Precedents and International Jurisprudence

The international debate surrounding AI inventorship was crystallized by the DABUS (Device for the Autonomous Bootstrapping of Unified Sentience) cases, spearheaded by Dr. Stephen Thaler and the Artificial Inventor Project.

Dr. Thaler filed two patent applications naming DABUS as the sole inventor:

1. A fractal beverage container based on fractal geometry.
2. A neural flame beacon for emergency signaling.

The global judicial response provides a comprehensive blueprint of modern international patent law:

┌───────────────────────────────────────────────────────────────────────────────┐
│                      THE GLOBAL DABUS LITIGATION MAP                          │
├─────────────────┬────────────────────────────┬────────────────────────────────┤
│ Jurisdiction    │ Forum / Authority          │ Outcome & Ruling               │
├─────────────────┼────────────────────────────┼────────────────────────────────┤
│ United States   │ USPTO; CAFC (*Thaler 2022*)│ REJECTED: Inventor must be an  │
│                 │                            │ individual natural person.     │
├─────────────────┼────────────────────────────┼────────────────────────────────┤
│ European Union  │ EPO Legal Board of Appeal  │ REJECTED: EPC Rule 19(1)       │
│                 │ (*J 0008/20, J 0009/20*)   │ mandates human inventor name.  │
├─────────────────┼────────────────────────────┼────────────────────────────────┤
│ United Kingdom  │ UK Supreme Court (2023)    │ REJECTED: Section 7 requires   │
│                 │ [*2023*] UKSC 49           │ natural person; no accession.  │
├─────────────────┼────────────────────────────┼────────────────────────────────┤
│ Australia       │ Full Federal Court (2022)  │ REJECTED: Reversed trial court;│
│                 │ [*2022*] FCAFC 62          │ Patents Act requires human.    │
├─────────────────┼────────────────────────────┼────────────────────────────────┤
│ South Africa    │ CIPC (2021)                │ GRANTED (Formal registration   │
│                 │                            │ system without prior exam).    │
└─────────────────┴────────────────────────────┴────────────────────────────────┘

A. United States: The Federal Circuit & USPTO Guidance

In Thaler v. Vidal (43 F.4th 1207, Fed. Cir. 2022), the United States Court of Appeals for the Federal Circuit held that under 35 U.S.C. § 100(f), an “inventor” is explicitly defined as “the individual or, if a joint invention, the individuals who invented or discovered the subject matter of the invention.”

The court relied on the plain linguistic meaning established by the US Supreme Court in Mohamad v. Palestinian Authority (2012), confirming that the word “individual” in federal statutes ordinarily refers to a natural human being unless Congress clearly indicates otherwise.

In February 2024, the USPTO issued its Comprehensive Inventorship Guidance for AI-Assisted Inventions:

• AI-assisted inventions are not categorically unpatentable.
• Patentability requires that a natural human person made a significant contribution to every claim.
• The USPTO adapted the landmark Pannu Factors (Pannu v. Iolab Corp., 1998) to evaluate human-AI collaboration:
  1. Did the human contribute in some significant manner to the conception or reduction to practice?
  2. Was the human contribution non-trivial relative to the overall invention?
  3. Did the human do more than merely present an existing problem or query to an AI system?

B. European Patent Office (EPO): Articles 58, 60, and 81 EPC

The EPO Legal Board of Appeal in J 0008/20 and J 0009/20 rejected the DABUS applications under Article 81 and Rule 19(1) EPC:

• Rule 19(1) EPC requires the designation of the inventor to state the family name, given names, and full address of the inventor.
• The Board confirmed that under the European Patent Convention, legislative provisions concerning inventorship confer moral and personal rights that presuppose legal personality. An AI machine has no legal capacity and cannot transfer rights to an applicant under Article 60(1) EPC.

C. United Kingdom: UK Supreme Court (Thaler v. Comptroller-General [2023] UKSC 49)

In December 2023, the UK Supreme Court unanimously dismissed Dr. Thaler’s appeal. Lord Kitchin articulated three decisive findings:

1. Under Section 7(3) of the Patents Act 1977, an “inventor” is strictly the natural person who devised the invention.
2. Dr. Thaler was not entitled to apply for a patent based on the ownership of DABUS.
3. The common-law doctrine of accession (where the owner of tangible property owns its fruits/yield, e.g., calves born to a cow) does not apply to intangible intellectual property concepts generated by software.

4. Substantive Patentability Challenges

Beyond inventorship attribution, AI creates deep structural friction with core patent doctrines:

A. The “Person Skilled in the Art” (PSITA / POSITA) Standard

The statutory test for an inventive step (Section 2(1)(ja) in India, Section 103 in the US, Article 56 EPC) asks whether the invention would have been obvious to a hypothetical Person Skilled in the Art.

As AI research assistants become ubiquitous standard tools in laboratories:

• The baseline capabilities of the PSITA are radically elevated.
• If a PSITA equipped with an AI tool can routinely screen millions of molecular compounds or simulate thousands of mechanical topologies in seconds, what remains truly “non-obvious”?
• The Paradox: Raising the PSITA standard to include standard AI tooling makes ordinary inventions unpatentable, while failing to do so allows companies to patent millions of routine AI-generated permutations.

Classical PSITA Threshold:
[Human Laboratory Researcher]  ──►  Standard Obviousness Bar

AI-Augmented PSITA Threshold:
[Human + High-Throughput AI Agent] ──►  Radically Higher Obviousness Bar

B. Sufficiency of Disclosure and the “Black Box” Problem

Under Section 10(4) of the Indian Patents Act and 35 U.S.C. § 112, a patent specification must describe the invention, its operation, and the best method known to enable any person skilled in the art to replicate it.

Deep learning systems (e.g., billion-parameter transformer models or neural networks) often suffer from the interpretability problem:

• An engineer may know the input training data and the output result, but cannot fully trace the latent neural weight interactions that arrived at the solution.
• If the model’s random seed, dynamic weights, or proprietary training corpus are not fully disclosed, the patent may fail the statutory mandate of enablement and sufficiency of disclosure.

5. Intersections with Geographical Indications (GI)

In intellectual property curricula, Patents and Geographical Indications represent contrasting philosophical poles:

Generative AI presents a distinct threat to GI ecosystems (such as Darjeeling Tea, Kanchipuram Silk, or Kashmiri Pashmina). Generative design tools can mimic distinctive regional visual patterns, textile motifs, and manufacturing aesthetics, producing cheap synthetic imitations that dilute the economic goodwill of registered GI communities.

Regulators are exploring whether AI training on datasets containing protected traditional knowledge and GI heritage requires prior informed consent under the Biological Diversity Act, 2002 and international Nagoya Protocol obligations.

6. Proposed Policy Reforms & The Road Ahead

To resolve the vacuum created by autonomous machine generation, scholars and legislative bodies have proposed three potential regulatory models:

┌─────────────────────────────────────────────────────────────┐
│                 PROPOSED REGULATORY MODELS                  │
├─────────────────────────────────────────────────────────────┤
│ 1. THE HUMAN STEWARDSHIP MODEL (Current Status Quo)         │
│    - Maintain human inventorship exclusively.               │
│    - Require human prompt engineering / parameter design    │
│      to qualify as significant inventive contribution.      │
├─────────────────────────────────────────────────────────────┤
│ 2. SUI GENERIS AI DISCOVERY PROTECTION                      │
│    - Short-term (e.g., 5–7 year) protection for autonomous │
│      AI inventions without full 20-year patent monopoly.    │
│    - Designed to incentivize compute investment while       │
│      preventing patent thickets.                            │
├─────────────────────────────────────────────────────────────┤
│ 3. STATUTORY EMPLOYER PRESUMPTION                           │
│    - Similar to UK Copyright Designs & Patents Act (CDPA)   │
│      Section 9(3) for computer-generated works:             │
│      "Inventor is the person by whom the arrangements       │
│      necessary for creation were undertaken."               │
└─────────────────────────────────────────────────────────────┘

7. Conclusion

The global consensus across India, the United States, Europe, and the United Kingdom remains resolute: an Artificial Intelligence system cannot be recognized as an inventor under current patent law.

However, inventions developed with the assistance of AI remain fully patentable, provided that human inventors demonstrate a significant, non-obvious intellectual contribution to the inventive step.

For engineers, students of patent law, and research institutions, the strategy is clear:

1. Document Human Ingenuity: Maintain meticulous laboratory notebooks detailing the formulation of the problem, parameter structuring, hypothesis pruning, and physical reduction to practice.
2. Ensure Complete Disclosure: Provide full architectural workflows, training parameters, and replication methodology to satisfy enablement requirements under Section 10(4).
3. Harmonize Legal & Engineering Workflows: Recognize that AI accelerates discovery, but human intellect remains the sole legal anchor of patentability.