IPR
Topic 90 AI Patents
Topic 90 — Artificial Intelligence and Patents
AI patentability operates at the intersection of two distinct issues: (i) Whether AI-generated inventions can be patented (with AI as inventor); and (ii) Whether AI-related inventions (algorithms, software architectures, machine learning systems) qualify as patentable subject matter under Section 3(k) of the Patents Act 1970. India's approach has crystallised through the Computer Related Inventions (CRI) Guidelines 2025, released on 29 July 2025 by the Indian Patent Office. The Guidelines apply with immediate effect and provide structured examination frameworks for AI, ML, Deep Learning, Blockchain, and Quantum Computing inventions. Critical Indian cases — Ferid Allani v. Union of India (2019:DHC:6944), Microsoft Technology Licensing v. Asst Controller (2023:DHC:3342), OpenTV v. Controller (2023:DHC:3305), Raytheon v. CGPDTM (2023:DHC:6673), Ab Initio Technology v. Controller (2024:DHC:5708), Blackberry v. Asst Controller (2024:DHC:6571), Comviva Technologies (2024), Google v. Controller (Del HC 2 April 2024) — have established the "technical effect" / "technical contribution" / "technical advancement" test. India is unique in maintaining ABSOLUTE business method exclusion (unlike UK and EP). Section 6 of the Patents Act requires a HUMAN inventor — globally consistent rejection of DABUS (Thaler) AI-as-inventor claims. This topic walks through the entire AI-patent framework — Section 3(k) interpretation, the CRI Guidelines 2025, the technical effect test, and AI inventorship.
1. Two Questions for AI and Patents
Q1 INVENTORSHIP Can AI be inventor? | Q2 SUBJECT MATTER AI software patentable? |
2. AI Inventorship — Section 6
A. Section 6 — Persons Entitled to Apply
Section 6 Patents Act 1970 "Persons entitled to apply for patents.— (1) Subject to the provisions contained in section 134, an application for a patent for an invention may be made by any of the following persons, that is to say,— (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 in respect of the right to make such an application; (c) by the legal representative of any deceased person who immediately before his death was entitled to make such an application." |
B. The "True and First Inventor" Requirement
- Section 6 requires "person" — implies natural person (human).
- "True and first inventor" — requires conception of invention.
- AI cannot conceive — operates from data and algorithms.
- AI cannot be assignee or legal representative.
- Courts globally have rejected AI as inventor.
✅ AI as inventor — globally rejected (Thaler/DABUS) Stephen Thaler's DABUS (Device for Autonomous Bootstrapping of Unified Sentience) system listed as inventor on patent applications globally: USA — USPTO rejected DABUS as inventor (2020); affirmed by Federal Circuit (2022); Supreme Court declined certiorari (2023). UK — UK IPO rejected; UK Court of Appeal affirmed (2021); UK Supreme Court affirmed (2023). EU — EPO rejected (2020); affirmed by EPO Board of Appeal (2021). Germany — Federal Patent Court rejected (2021). Australia — Federal Court initially recognised (2021); then reversed by Full Court (2022). India — DABUS applications also rejected. CRI Guidelines 2025 explicitly state: "AI-generated inventions, produced autonomously by an AI system with minimal human intervention, are NOT patentable: section 6 of the Patents Act requires a human inventor." The global consensus: AI cannot be inventor. |
C. AI-Assisted vs AI-Generated
Category | Description | Patentability |
|---|---|---|
AI-assisted invention | Human inventor; AI as tool helping create. | PATENTABLE — human is "true and first inventor". |
AI co-invention | Human + AI both contribute. | PATENTABLE if human contribution sufficient. |
AI-generated invention | Pure AI generation; minimal human input. | NOT PATENTABLE — no human inventor. |
3. Section 3(k) — AI Software Patentability
A. The Statutory Framework
Section 3(k) Patents Act "What are not inventions.— (k) a mathematical or business method or a computer programme per se or algorithms;" |
B. The "Per Se" Qualifier
Section 3(k) excludes computer programs PER SE (alone, in isolation):
- "Per se" added by Patents (Amendment) Act 2002 (effective 20 May 2003).
- Means: pure software cannot be patented.
- But software with technical effect/contribution may be patented.
- Indian position between EU (more permissive) and traditional restrictive view.
C. The Technical Effect/Contribution/Advancement Test
📖 Ferid Allani v. Union of India & Ors., 2019:DHC:6944 (Del HC) Facts — Patent application rejected on grounds of Section 3(k) exclusion. Petitioner challenged. Holding — Delhi HC established the foundational test: "Computer programs must be assessed for their technical contribution, technical effect, and technical advancement to be considered patentable." Three-prong test: (i) Technical contribution — solves a technical problem. (ii) Technical effect — produces concrete technical outcome. (iii) Technical advancement — improves on prior art technically. Significance — Foundational Indian case for software patentability. Cited in nearly every subsequent CRI case. Ended the rigid "per se" exclusion approach. Adopted in CRI Guidelines 2025. |
4. CRI Guidelines 2025 — Detailed Framework
A. AI/ML Examination Approach
✅ CRI Guidelines 2025 — AI/ML Patentability For AI-related inventions to be patentable: · Must demonstrate practical application providing technical solution. · Must transform abstract AI principle into real-world application. · Must not be merely theoretical model. · Must not be solely abstract algorithm. Examples PATENTABLE: · AI system controlling industrial robotic arm. · Machine learning-based image processing for medical diagnostics. · Neural network for noise reduction in communications. · Computer vision for manufacturing defect detection. · Speech recognition with specific signal processing. Examples NOT PATENTABLE: · Pure mathematical AI algorithm. · Generic data classification. · Business method using AI. · AI-driven market prediction (business method). · Theoretical optimization model. The CRI Guidelines emphasize: · Sufficient disclosure required for AI/ML. · Training data and parameters specified where relevant. · Concrete implementation in technical field. |
B. The Three-Step Test
- Understand the invention as a whole.
- Determine whether it merely reflects a commercial strategy or instead provides a technical solution.
- Rule on patentability.
C. Mathematical Method Exclusions
CRI Guidelines 2025 clarify mathematical method exception:
- Pure mathematical methods — NOT patentable.
- Mathematical formulae in encoding/decoding systems — MAY be patentable.
- Noise reduction in communications using mathematical methods — MAY be patentable.
- Encryption/decryption algorithms with technical implementation — MAY be patentable.
- Electrical/electronic systems using mathematical methods — MAY be patentable.
5. Key Indian AI/Software Patent Cases
📖 Microsoft Technology Licensing LLC v. Asst Controller of Patents, 2023:DHC:3342 (Del HC) Facts — Microsoft software patent application; Section 3(k) rejection appealed. Holding — Delhi HC established framework: (i) Software patents not categorically excluded. (ii) Technical effect demonstration crucial. (iii) Detailed examination of technical contribution required. (iv) Examination must consider invention as a whole. Significance — Reinforced Ferid Allani framework; Microsoft's software patent grounds. |
📖 OpenTV Inc. v. Controller of Patents and Designs, 2023:DHC:3305 (Del HC) Facts — OpenTV patent application for TV-related software/business method. Holding — Delhi HC clarified BUSINESS METHOD bar: (i) Business methods are ABSOLUTELY excluded under Section 3(k). (ii) Even technically implemented business methods excluded. (iii) Substance over form — if invention is fundamentally a business solution, not patentable. (iv) India's position differs from UK/EU which permit technically implemented business methods. Significance — Confirmed India's ABSOLUTE business method exclusion. Critical distinction from UK/EU/US. Fintech, e-commerce, digital business platforms face significant Indian patent restrictions. |
📖 Raytheon Company v. Controller General of Patents and Designs, 2023:DHC:6673 (Del HC) Facts — Raytheon patent application; Section 3(k) issues. Holding — Delhi HC examined CRI examination criteria. Reinforced technical effect requirement. Detailed analysis of computer-implemented inventions. Significance — Cited in CRI Guidelines 2025; standards for technical disclosure. |
📖 Ab Initio Technology LLC v. Controller of Patents & Designs, 2024:DHC:5708 (Del HC) Facts — Patent application related to BLOCKCHAIN technology rejected on Section 3(k) grounds. Holding — Delhi HC clarified: "If an innovative input in the form of a process, system, or method enhances the computational ability of the processor, the effect is technical." Blockchain inventions: (i) Theoretical concept of distributed ledger — NOT patentable. (ii) Practical implementation with concrete technical outcome — MAY be patentable. (iii) Technical effect must be specific (latency reduction, tamper-proof provenance, etc.). Significance — Foundational case for BLOCKCHAIN patents in India. Established that blockchain not categorically excluded if technical effect demonstrable. |
📖 Comviva Technologies Limited v. Asst Controller of Patents, 2024 (Del HC) Facts — Comviva patent application for "Methods and Devices for Authentication of Electronic Payment Card using Electronic Token" (IN201611000234). Initially rejected as business method + computer programme per se. Holding — Delhi HC observed: (i) Business method involvement does not render invention unpatentable per se. (ii) Technical features that enhance security in electronic transactions — patentable subject matter. (iii) Mere involvement of business activity does not preclude patent. (iv) Technical solution to technical problem — patentable. Order — Remanded for reconsideration; instructing assessment of technical merits beyond business aspects. Significance — Important fintech/payment patent case. Confirmed: technical features within business context can be patented. |
📖 Google Inc. v. Controller of Patents (Del HC, 2 April 2024) Facts — Google's Indian Patent Application No. 5429/DELNP/2007 for managing instant messaging sessions across multiple devices. Rejected by IPO in 2019 for lack of inventive step under Section 2(1)(ja). Holding — Delhi HC found: (i) Invention provided technical advancement over existing solutions. (ii) Set aside IPO's rejection. (iii) Directed grant of patent. Significance — Software patent successfully granted. Demonstrates importance of demonstrating technical advancement. |
📖 Blackberry v. Asst Controller of Patents and Designs, 2024:DHC:6571 (Del HC) Facts — Blackberry patent application; CRI examination standards. Holding — Delhi HC reinforced examination standards. Detailed framework for technical effect demonstration. Significance — Cited in CRI Guidelines 2025. |
📖 Telefonaktiebolaget LM Ericsson v. Lava International Ltd., 2024:DHC:2698 (Del HC) Facts — Major patent infringement case involving telecom standards. Holding — Delhi HC established SEVEN-STEP NOVELTY TEST: (1) Identify the invention. (2) Identify the prior art. (3) Compare claim by claim. (4) Assess differences. (5) Identify inventive step. (6) Apply person skilled in art test. (7) Conclude on novelty. Significance — Adopted by CRI Guidelines 2025 for novelty assessment. Standard novelty framework. |
6. AI/ML Specific Considerations
A. Disclosure Requirements
CRI Guidelines 2025 require detailed disclosure for AI/ML inventions:
- Training data sources and characteristics.
- Model architecture (if specific).
- Training methodology.
- Hyperparameters where relevant.
- Validation and testing procedures.
- Performance metrics demonstrating technical effect.
- Hardware/software platform requirements.
B. Industrial Application
Application Domain | Patentability |
|---|---|
Medical AI (diagnostics, treatment) | Patentable if technical solution to medical problem. |
Autonomous vehicles | Patentable for technical features (sensors, control). |
Image/video processing | Patentable for technical processing. |
Speech recognition | Patentable with specific signal processing. |
Robotics | Patentable for technical control systems. |
Manufacturing AI | Patentable for technical process improvements. |
Financial AI | NOT patentable (business method exclusion). |
Marketing AI | NOT patentable (business method). |
HR AI | NOT patentable (business method). |
7. International AI Patent Comparisons
Jurisdiction | AI Inventorship | Software/AI Patentability |
|---|---|---|
India | Section 6 — human required; AI-generated NOT patentable. | Section 3(k); technical effect test; CRI Guidelines 2025. |
USA | Inventorship requires "natural person" (Thaler v. Vidal 2022). | Alice/Mayo two-step test; abstract idea exclusion. |
EU/EPO | "Human inventor" required (DABUS rejected). | Technical character test; patentability if technical contribution. |
UK | Designating AI as inventor not permitted (Thaler 2023 SC affirmed). | Technical contribution test; software with technical effect patentable. |
China | Initially rejected DABUS; some state recognition emerging. | More permissive; AI inventions widely patented. |
Japan | Human inventor required; DABUS rejected. | Software patentable if produces technical effect. |
8. Strategic Considerations
✅ For AI patent applicants — twelve points Identify HUMAN inventors clearly — never AI as inventor. Document human creative role in AI-assisted invention. Demonstrate technical effect prominently — Ferid Allani three-prong test. For AI/ML, comprehensive disclosure: training data, architecture, methodology. For blockchain, document technical implementation beyond theoretical concept (Ab Initio). For business-related AI, recognize ABSOLUTE business method exclusion (OpenTV). For payment authentication or similar, focus on technical security improvements (Comviva). For software patents, both system AND method claims if supported by description. For novelty, apply seven-step test (Ericsson v. Lava). For inventive step, demonstrate technical advancement. For specifications, fully disclose with best method (Section 10). For litigation strategy, monitor pending cases like Comviva continuation. |
🎯 EXAM POINTERS — TOPIC 90
|