Artificial intelligence and software are transforming every sector of the economy – from drug discovery and autonomous vehicles to financial modeling and legal research. As these technologies generate increasingly sophisticated outputs, the question of whether they can be protected by patents has become one of the most contested and consequential debates in intellectual property law. In India, this debate is anchored in a statutory framework that was drafted in an era when software was in its infancy and artificial intelligence was a theoretical concept. The challenge for practitioners, applicants and courts is to interpret that framework intelligently in light of technological realities that its drafters could not have foreseen.
The short answer under Indian law is that software and AI inventions are not categorically excluded from patent protection – but they are not straightforwardly patentable either. The outcome depends critically on how the invention is framed, what technical contribution it makes, and how its claims are drafted. Understanding where the line falls requires a close examination of Section 3(k) of the Patents Act, 1970, the Guidelines for Examination of Computer Related Inventions issued by the Indian Patent Office, and the evolving judicial and administrative practice that has shaped this area.
The Statutory Exclusion – Section 3(k)
Section 3(k) of the Patents Act, 1970 provides that a mathematical method, a business method, a computer programme per se or algorithms shall not be regarded as an invention for the purposes of the Act. The operative phrase is “computer programme per se.” This qualification is not accidental. Parliament did not exclude computer programmes absolutely. It excluded them only when claimed in their bare, abstract form – as software divorced from any technical application or effect.
The word “per se” creates the critical opening. It signals that a computer programme embedded within a larger invention – one that produces a technical effect or solves a technical problem – may qualify as patentable subject matter, even though a bare algorithm standing alone would not. The drafting reflects a deliberate policy choice to exclude abstract intellectual constructs while preserving the possibility of patent protection for inventions with genuine technical character.
The same logic applies to mathematical methods and algorithms. A mathematical formula in isolation is not patentable. However, a mathematical method implemented in a specific technical context – such as a signal-processing algorithm that improves the efficiency of data transmission, or a machine-learning model that diagnoses disease from medical imaging – may cross the threshold if it produces a concrete technical result.
The CRI Guidelines – The Indian Patent Office’s Position
The Indian Patent Office has issued Guidelines for Examination of Computer Related Inventions, most recently updated in 2017, to assist examiners in assessing applications involving software, algorithms and related technologies. These guidelines remain the primary administrative instrument governing examination practice in this area and are essential reading for any practitioner filing AI or software patent applications in India.
The CRI Guidelines adopt a three-stage analytical approach. First, the examiner identifies the actual contribution of the claimed invention. Second, the examiner assesses whether that contribution is solely in the excluded categories – mathematical methods, business methods, computer programmes per se or algorithms. Third, if the contribution lies solely in the excluded area, the application is rejected under Section 3(k). If the contribution extends beyond those categories to produce a technical effect, the application may proceed.
The Guidelines make clear that the mere fact that a computer or processor is involved does not save an invention from Section 3(k) exclusion. Nor does the presence of hardware in the claim. What matters is whether the combination of software and hardware produces a technical effect that goes beyond the normal physical interactions between a programme and the computer on which it runs. Examples of qualifying technical effects cited in the Guidelines include faster processing, reduced memory usage, more efficient data storage, improved signal transmission and enhanced system performance.
The CRI Guidelines introduced what is sometimes called the “technical effect” test – borrowed in part from European Patent Office jurisprudence – as the central criterion for distinguishing patentable computer-implemented inventions from excluded software per se. This approach has significantly shaped examination practice, though its application remains uneven and fact-specific.
Artificial Intelligence Inventions – Specific Challenges
Artificial intelligence inventions present the Section 3(k) challenge in its most acute form. A machine learning model is, at its core, a mathematical construct – a function that maps inputs to outputs through parameters derived from training data. A neural network is a system of weighted mathematical operations. At the level of abstraction, AI is mathematics. The question is whether a patent claim directed to an AI invention can be framed in a manner that transcends this mathematical foundation and captures a technical contribution.
Indian patent law has not yet produced authoritative judicial guidance specifically addressing AI inventions. However, several principles from general CRI examination practice apply directly.
An AI invention is most likely to survive Section 3(k) scrutiny when the claim is directed to a specific technical application rather than to the AI model or algorithm in the abstract. A patent claim directed to “a method of training a neural network” is far more vulnerable than one directed to “a medical imaging system comprising a convolution neural network configured to identify malignant tumors from CT scan data with a specificity of at least ninety percent.” The second formulation ties the AI component to a concrete technical problem in a defined field and specifies a measurable technical result.
The distinction between AI as a tool and AI as the invention is critical. Where AI is deployed as a means to achieve a result in a physical, technical system – controlling the dynamics of a vehicle, optimizing the operation of an industrial process, diagnosing a condition from sensor data – the invention has a technical character that extends beyond the algorithm itself. Where the AI is the subject of the claim without connection to a specific technical application, Section 3(k) presents a substantial barrier.
Claim Drafting Strategy – How to Navigate Section 3(k)
The fate of a software or AI patent application in India is determined substantially at the drafting stage. Claim drafting is not merely a technical exercise – it is a legal strategy for placing the invention on the right side of the Section 3(k) line.
Several principles guide effective drafting for this category of inventions. Claims should be framed as systems or methods rather than as programmes or algorithms in isolation. A claim to “a computer-implemented method for optimizing power grid load distribution comprising the steps of…” is structurally more defensible than a claim to “an algorithm for optimizing power distribution.” The former situates the software within a technical process with an identifiable field of application.
The technical effect should be expressly identified in the claim or at minimum in the specification with a clear link to the claimed steps. Generic statements that the invention “improves efficiency” or “enhances performance” are unlikely to satisfy an examiner. Specific, measurable technical effects – reduction in computational latency by a defined percentage, improvement in classification accuracy over prior art systems, reduction in power consumption – provide the evidentiary basis for distinguishing the contribution from a mere algorithm.
Hardware integration strengthens a claim. Where the software or AI model operates in conjunction with specific hardware components – sensors, processors configured in a particular architecture, communication interfaces – the claim reflects a technical system rather than an abstract process. The interaction between software and hardware, producing a defined technical outcome, is the essence of a patentable computer-implemented invention under Indian law.
Method claims and system claims should both be pursued. A method claim captures the process of using the invention. A system claim captures the apparatus that implements it. Together they provide complementary scope and reduce the risk that a narrow interpretation of one claim type defeats the entire application.
The Interplay with Other Patentability Requirements
Section 3(k) is the primary but not the only obstacle for AI and software patent applications. An application that successfully navigates the subject matter exclusion must still satisfy the standard patentability requirements of novelty under Section 2(1)(l), inventive step under Section 2(1)(ja) and industrial applicability under Section 2(1)(ac).
The inventive step analysis for AI inventions is particularly demanding. The field of AI research is extraordinarily active, with thousands of papers published annually describing new architectures, training methods and applications. Prior art in this space is dense and evolving rapidly. An AI invention that appears novel at the time of filing may face significant obviousness challenges if the claimed advance represents a straightforward application of known techniques to a new domain – which examiners frequently argue is the case.
Industrial applicability is generally less problematic for AI inventions that are directed to real-world technical applications, but it remains relevant where the claimed invention operates at a level of abstraction that makes its practical utility unclear. An AI model for “general intelligence enhancement” would likely fail industrial applicability requirements for want of a defined technical application.
International Context – A Comparative Perspective
Understanding India’s approach benefits from comparison with other major patent systems. The European Patent Office similarly excludes mathematical methods, mental acts and computer programmes as such, but permits computer-implemented inventions where a technical character is demonstrated. EPO practice has developed a sophisticated jurisprudence on what constitutes a technical effect, and Indian examination practice has drawn substantially from this framework, as reflected in the CRI Guidelines.
The United States, following the Supreme Court’s decision in Alice Corporation v. CLS Bank International (2014), applies a two-step abstract idea analysis that has resulted in widespread rejection of software and AI claims framed at a high level of generality. However, US practice has evolved to permit AI claims directed to specific technical improvements, and the USPTO has issued guidance on AI patentability that is more permissive than Alice alone might suggest.
China has become one of the most significant jurisdictions for AI patent filings globally, with a rapidly growing body of granted patents in machine learning, computer vision and natural language processing. Chinese patent examination guidelines specifically address AI inventions and permit protection where a technical solution to a technical problem is demonstrated.
India’s framework is broadly aligned with the European technical character approach, though examination practice has not yet achieved the same level of doctrinal sophistication. As AI patent applications increase in volume before the Indian Patent Office, more developed examination standards and eventually judicial guidance are inevitable.
Who Is Filing AI and Software Patents in India – and What This Means
The volume of computer-related invention filings before the Indian Patent Office has grown substantially over the past decade. Applicants include multinational technology companies, Indian software firms, pharmaceutical companies deploying AI in drug discovery, automotive manufacturers developing autonomous systems, and, increasingly, academic and government research institutions.
The profile of applicants matters for examination practice. A well-resourced multinational filing with detailed claim sets, thorough prior art analysis and experienced patent counsel is better positioned to navigate examination objections than an individual inventor or startup filing without specialist support. The CRI examination process frequently involves multiple rounds of objections and responses. Applicants should anticipate Section 3(k) objections as a near-certainty and prepare substantive arguments addressing the technical effect of the invention before the First Examination Report is even issued.
For Indian startups and technology companies in particular, the cost and complexity of AI patent prosecution represents a real barrier. The National IPR Policy, 2016 and various startup IP facilitation schemes offer fee concessions and expedited examination for eligible applicants. Startups should explore these mechanisms early in the filing process. The Startup India IP Protection scheme, administered through the Controller General of Patents, provides access to panel facilitators who can assist with prosecution at subsidized rates.
Conclusion
The patentability of AI and software inventions in India is neither a closed door nor an open one. Section 3(k) of the Patents Act, 1970 excludes computer programmes per se and algorithms, but the qualifier “per se” preserves meaningful space for inventions that produce a technical effect beyond the abstract execution of software. The CRI Guidelines operationalize this space through the technical effect test, directing examiners to assess the actual contribution of the claimed invention rather than its form.
For AI inventions specifically, the critical strategic imperative is to anchor claims in concrete technical applications, specify measurable technical effects, integrate hardware where relevant and avoid framing the invention at the level of the algorithm or mathematical model alone. An AI invention claimed as a technical system solving a defined technical problem in a specific field has a substantially greater prospect of success than one claimed in abstract terms.
As AI continues to penetrate every domain of technology and commerce, the legal framework governing its patentability will face increasing pressure to develop. India’s growing position as a technology innovator – in software services, pharmaceuticals, space technology and beyond – makes the resolution of these questions not merely a matter of intellectual property doctrine but of national innovation policy.
References
- The Patents Act, 1970 – https://ipindia.gov.in
- Guidelines for Examination of Computer Related Inventions, 2017 – https://ipindia.gov.in/writereaddata/Portal/IPOGuidelinesManuals/1_73_1_CRI_Guidelines_August_2015.pdf
- Manual of Patent Office Practice and Procedure – https://ipindia.gov.in
- Alice Corporation v. CLS Bank International, 573 U.S. 208 (2014)
- EPO Guidelines for Examination – https://www.epo.org/law-practice/legal-texts/guidelines.html
- TRIPS Agreement – https://www.wto.org/english/docs_e/legal_e/27-trips.pdf
- National IPR Policy, 2016 – https://dpiit.gov.in
Compulsory Licensing Copyright Act Copyright Act 1957 Copyright Infringement Copyright Law Copyright Rules Deceptive Similarity Goodwill Indian IP Framework Indian Patent Law Indian Trademark Law Intellectual Property Law IP Law India Passing Off Patent Claims Patent Enforcement Patent Infringement Patent law Patent Revocation Patent Rule Pharmaceutical Patents Section 3 Section 29 The Patent Act 1970 Trademark Enforcement Trademark Registration Trade Marks Act 1999 Trade Marks Rules 2017 TRIPS Compliance