01 — The Core Obstacle
The Section 3(k) Challenge
The Patents Act, 1970 grants protection to inventions that are novel, involve an inventive step and are industrially applicable. But it also draws boundaries to keep abstract ideas out of the patent system.
Section 3(k) · The exclusion
Mathematical methods, business methods, computer programs "per se" and algorithms are not patentable inventions.
Since AI and ML are fundamentally algorithm-driven, this clause is the central concern. The crucial word, though, is per se. The Indian Patent Office — guided by the Guidelines for Examination of Computer-Related Inventions (CRIs) — has read the exclusion to cover standalone software, not inventions that produce a genuine technical effect or contribution.
Why AI/ML isn't just "Software"
01 Learns from Data
It improves over time from data, without explicit reprogramming by a human.
02 Decides Autonomously
It classifies, predicts and makes decisions — sometimes beyond human capability.
03 Evolves Dynamically
Its adaptive, evolving nature resists a clean label of "abstract idea."
That distinction is exactly what gives applicants room to argue an AI/ML invention is more than a computer program "per se."
02 — Clearing the Bar
Eligibility & Technical Effect
An AI/ML invention becomes patent-eligible when it does more than run an algorithm — when it is tied to hardware or solves a technical problem in a technical field. That "technical effect" is the heart of the case for grant.
What Counts as a Technical Effect
Improved Efficiency
Reduced system latency, faster real-time predictions, lower power consumption.
Enhanced Performance
Higher accuracy on complex tasks beyond what prior methods achieve.
Real-world Impact
Tangible industrial benefit — predictive analytics in manufacturing, for example.
Concrete examples that tend to read as technical: AI improving real-time image processing in autonomous vehicles, an ML model optimising pharmaceutical compound discovery, or a method that cuts power use in server systems. Alongside the technical effect, the invention must still show technical advancement over what exists and industrial applicability in a real commercial setting.
03 — The Harder Test
Novelty & Inventive Step
Prior art in AI is broad — it includes existing patents, published research, and publicly available code, so even an open-source repository on GitHub can be cited to challenge novelty. The tougher question is usually inventive step: would the invention be obvious to a person skilled in the art?
Likely patentable
Solving a Technical Problem
AI applied in a non-obvious way that overcomes a genuine technical limitation.
Likely not
Obvious Automation
Simply automating a known manual process with AI, with no unexpected technical gain.
04 — The Lifecycle
The Patent Process in India
The procedure mirrors any Indian patent filing — with the examiner's attention on Section 3(k) at the objection stage.
1 Prior Art Search
Novelty check
Search global databases — and code repositories — to confirm the invention is new.
2 Draft the Application
Frame the technical contribution
Articulate the technical problem and the novel solution, with diagrams and data-flow detail.
3 File with the IPO
Forms & fees
Submit the application with the required forms and fees to the Indian Patent Office.
4 Publication
18 months
The application is published after 18 months from priority, unless early publication is sought.
5 Request Examination
Within 48 months
File the examination request; the IPO assesses the application for patentability.
6 Respond to Objections
Address Section 3(k)
Reply to the examiner's report — most critically, the arguments on the 3(k) exclusion.
7 Grant of Patent
20-year term
On meeting all requirements, the patent is granted and published — protection for 20 years from filing.
05 — What Works, What Fails
Pitfalls & Best Practices
Common pitfalls
- Over-reliance on vague algorithm descriptions
- No demonstration of real-world applicability
- Broad claims that don't distinguish from prior art
- Treating the AI model as the invention, not its effect
Best practices
- Lead with the technical problem and novel solution
- Link the software to specific hardware where possible
- Use clear diagrams, workflows and data-flow explanations
- Anchor claims to a measurable technical effect
Where Grants have Happened
AI/ML patents have been granted in India where technical advancement and practical utility were clear — for example, medical-diagnostics platforms with improved accuracy, ML fraud-detection systems in banking, and smart energy-optimisation solutions combining IoT and ML.
06 — The Wider Toolkit
Beyond Patents: A Layered IP Strategy
Patents are one instrument. A robust AI/ML portfolio usually combines several forms of protection.
COPYRIGHT
Code & Generated Works
Source code and certain AI-assisted works can be protected under the Copyright Act, 1957 — though only human-authored works qualify, and authorship of AI-generated output turns on human control and input.
TRADE SECRETS
Models & Datasets
Proprietary models, weights and training data are often best kept as trade secrets. India has no dedicated statute, so protection rests on contracts and common law — NDAs and access controls.
DESIGN
Interface & UI/UX
A novel graphical interface for an AI tool may qualify for protection under the Designs Act, 2000, provided it is new, original and not previously disclosed.
SEMICONDUCTOR
Chip Layouts
Layout designs of integrated circuits used in AI hardware can be protected under the Semiconductor Integrated Circuits Layout-Design Act, 2000.
07 — Thinking Globally
Global strategy & government support
India is a signatory to the Patent Cooperation Treaty (PCT), so a single international application can reserve rights across member states — particularly valuable for startups eyeing global markets, with India as the base country.
Initiatives worth using
RESEARCH
National AI Mission
Promotes AI research and development across sectors.
STARTUPS
Startup India (SIPP)
The IP Protection scheme offers incentives and facilitator support for startups.
SUBSIDY
Patent facilitators
Provide subsidies and assistance for patent filing.
08 — In Summary
Conclusion
India's patent system is adapting to AI and ML, but the burden is on the applicant to show the invention is more than an algorithm. Understanding the Section 3(k) exclusion, demonstrating a real technical effect, avoiding the common drafting traps, and layering patents with trade secrets and design protection is what turns a breakthrough into a defensible asset.