Quick Answer
Yes, AI-related inventions can be patented in India, provided they are structured to overcome the statutory hurdles of Section 3(k) of the Patents Act, 1970. Pure mathematical methods, abstract algorithms, and computer programs per se are statutorily non-patentable. However, under the Guidelines for Examination of Computer-Related Inventions (CRIs) and prevailing High Court jurisprudence (such as Ferid Allani v. Union of India), if an artificial intelligence model or machine learning architecture produces a further technical effect or solves a tangible technical problem by interacting with hardware components, the invention qualifies for patent grant. Furthermore, under Section 6 of the Act, an AI entity cannot be recognized as an inventor; a natural person must be named as the true and first inventor.
Key Takeaways
- The "Technical Effect" Test: Abstract neural network topologies are unpatentable; applied AI solving concrete physical or computational bottlenecks (e.g., medical image denoising or battery optimization) is patentable.
- Ferid Allani Doctrine: The Delhi High Court established that the bar on software patentability is not absolute; if the software innovation demonstrates technical contribution, Section 3(k) does not apply.
- Inventorship Must Be Human: Consistent with global standards, Indian patent law does not permit artificial intelligence systems (like DABUS) to be named as inventors; inventorship requires natural human authorship.
- Hardware Interdependence: Claims must be drafted as an integrated system or apparatus, clearly tying algorithmic execution to physical processors, sensors, memory buffers, or control units.
Navigating Section 3(k) for AI Inventions
Section 3(k) bars the following four related categories:
- A mathematical method
- A business method
- A computer programme per se
- Algorithms
Because artificial intelligence models are fundamentally mathematical algorithms implemented via software, examiners routinely issue Section 3(k) rejections during FER examination. To successfully traverse these objections, patent drafting counsel must establish that the AI application creates an effect beyond ordinary computer execution:
- Internal Technical Effect: Increasing processing speed, reducing cache latency, improving data transmission security, or optimizing GPU memory allocations.
- External Technical Effect: Real-time control of an industrial robotic arm, automated steering of an autonomous vehicle, or computer-aided detection of oncological lesions from radiological scans.
Strategic Rules for Drafting AI Patent Claims
- Avoid Pure Algorithmic Descriptions: Do not frame claims around mathematical equations or training loss minimization in the abstract. Always anchor the model to specific input data structures and tangible output actions.
- Detail the Pre-Processing & Feature Extraction: The Patent Office requires full disclosure of how raw physical data is transformed into feature vectors before ingestion by the neural network.
- Include System & Apparatus Claims: Formulate claims as a "Computer-implemented system for [technical application], comprising: one or more processors coupled to memory..." rather than a standalone method claim.
- Provide Exhaustive Flowcharts: Specifications must include detailed flow diagrams showing exact data transitions between computational stages to satisfy Section 10(4) sufficiency requirements.
Patent Your Artificial Intelligence Innovations
Our patent attorneys possess specialized computer science backgrounds, ensuring your deep-tech AI and machine learning claims satisfy Indian CRI criteria and international PCT requirements.
Primary Legal Authorities & Sources
- The Patents Act, 1970 (Section 3(k), Section 6, and Section 10).
- Ferid Allani v. Union of India & Ors., 2019 SCC OnLine Del 11867.
- Microsoft Technology Licensing, LLC v. Assistant Controller of Patents, 2023 SCC OnLine Del 2772.
- Guidelines for Examination of Computer-Related Inventions (CRIs), CGPDTM, Government of India.