Understanding AI-based Inventions and Their Growing Relevance
Artificial Intelligence (AI) based inventions encompass technologies that empower machines and systems to emulate aspects of human intelligence such as learning from data, reasoning, prediction, decision-making, and pattern recognition through sophisticated computational models. Typically grounded in machine learning techniques, neural networks, natural language processing, computer vision, and autonomous architectures, these inventions are increasingly embedded across diverse sectors including healthcare, finance, manufacturing, telecommunications, and consumer technology.
The growing prominence of AI-based inventions is reflected in the sharp rise of AI-related patent filings worldwide, with offices such as the USPTO, EPO, and CNIPA reporting sustained growth in AI-driven applications. India has followed suit, fuelled by increased R&D activity, an expanding startup ecosystem, and policy initiatives such as Digital India and the IndiaAI Mission. As AI technologies become integral to core technological infrastructure, it brings patents squarely into focus as the principal vehicle for IP protection. Yet, AI-based inventions sit uneasily within traditional patent frameworks, given their algorithmic foundations, reliance on data-driven learning, and predominantly software-based implementation. The friction is particularly pronounced in India, where statutory exclusions for computer programs and algorithms continue to play a decisive role in determining patent eligibility.
AI-based Inventions Through the Lens of Intellectual Property and Patents
From an intellectual property perspective, artificial intelligence unsettles some of the most foundational assumptions of patent law. When innovation is driven by self-learning models and data-trained systems, can algorithms remain outside the scope of protection? Who qualifies as the ‘inventor’ when human ingenuity is interwoven with machine-generated outcomes? How should ‘technical contribution’ be evaluated when advancement lies not in hardware, but in improved data processing, model architecture, or adaptive optimisation?
Patent systems face a strain under the rise of AI, as they demand clearly defined thresholds of patentable subject matter, novelty, inventive step, and industrial applicability, standards that were conceived for human-devised, deterministic inventions. AI thus forces patent systems to confront a fundamental question: how to protect genuine technological innovation without diluting the statutory boundaries of exclusivity.
In India, these questions are predominantly addressed within the framework of the Patents Act, 1970, most notably through Section 3(k), which excludes from patentability ‘a mathematical or business method or a computer program per se or algorithms.’ Given that a vast majority of AI-based inventions are realised through software-driven architectures, patent applicants routinely face objections alleging that their claims fall within this statutory exclusion. As a result, the development of India’s patent regime for AI-based inventions has been shaped largely by the evolving interpretation and application of Section 3(k), seeking to strike a careful balance between fostering technological innovation and preserving the legislative intent underlying subject-matter exclusions.
The Legal Journey: Section 3(k), CRI Guidelines, and Judicial Interpretation
India’s approach to AI regulation has largely favoured innovation-first policy making. Rather than enacting AI-specific patent legislation, the Government has focused on adapting existing legal frameworks to accommodate emerging technologies. Policy initiatives by MeitY and DPIIT reflect an acknowledgment that AI will play a transformative role in economic growth, public services, and technological self-reliance.
At the policy level, parliamentary committees and expert bodies have recognised that traditional IP concepts may require re-examination in the context of AI. Discussions around AI-generated inventions, inventorship norms, and ownership structures have gained visibility, signalling potential long-term reforms. While no legislative amendments to the Patents Act have yet been introduced specifically for AI, the policy momentum clearly points towards enabling, rather than restricting, AI-driven innovation
Section 3(k): Statutory Exclusion and Its Evolving Interpretation
Section 3(k) of the Patents Act, 1970 excludes from patentability ‘a mathematical or business method or a computer program per se or algorithms.’ For over a decade, this provision has been the focal point of debate for software-implemented and AI-based inventions. Early examination practice tended to adopt a literal reading of the exclusion, often rejecting applications merely because claims involved software or algorithmic steps. However, Indian jurisprudence has gradually moved towards a purposive and substance-based interpretation.
A watershed moment in the interpretation of Section 3(k) came with the Delhi High Court’s decision in Ferid Allani v. Union of India & Ors. (2019), where the Court clarified that the statutory exclusion of ‘computer programs per se’ does not categorically bar all computer-related inventions from patentability. The Court observed that in the modern digital era, virtually all technologies including artificial intelligence and blockchain rely on computer programs, and it would be regressive to reject patents solely on this basis. What matters instead is whether the claimed invention, when considered as a whole, demonstrates a technical effect or technical contribution beyond a mere program, such that it solves a technical problem or produces a concrete technical benefit. Accordingly, the Court directed that patent applications involving software-implemented inventions should be re-examined with a focus on technical contribution rather than dismissed solely because they involve computer programs.
In Telefonaktiebolaget LM Ericsson v. Intex Technologies (India) Ltd., the High Court dealt with challenges under Sections 3(k) and 3(m) of the Patents Act to patents asserted by Ericsson in the telecommunications domain. The Court observed that the statutory exclusion of ‘computer programs per se’ under Section 3(k) does not extend to inventions that embody a technical contribution or technical effect, and that an invention implemented through software should not be considered unpatentable merely because it involves computer programs. In this context, the Court held that inventions which produce a demonstrable technical improvement in technology and are not abstract algorithms or mere computer programs per se fall outside the scope of the statutory exclusion and may be patentable. Although the case arose primarily in the context of standard essential patent enforcement rather than patent examination, the underlying principle that technical contribution and real-world technical effect prevail over mere software form has broad relevance to AI-driven systems, where innovations often lie in improved data processing, system optimisation, and functional enhancements rather than in abstract computational logic.
CRI Guidelines 2025: Institutionalising the Technical Effect Test
The Computer Related Inventions (CRI) Guidelines have played a crucial role in operationalising judicial principles at the examination level. The 2025 CRI Guidelines explicitly recognise emerging technologies such as artificial intelligence, machine learning, deep learning, and data-driven systems. They reinforce a stepwise analysis to determine whether the claimed subject matter is directed merely to an excluded category or whether it forms part of a larger technical solution.
The Guidelines clarify that the mere use of algorithms, mathematical models, or AI techniques does not attract the Section 3(k) exclusion where such elements operate as intermediate steps within a larger technical solution. What is determinative is whether the claimed invention, when assessed as a whole, produces a demonstrable technical effect such as improved processing speed, reduced latency, enhanced accuracy, efficient resource utilisation, improved image or signal quality, or real-time control of physical or technical systems. At the same time, the Guidelines make clear that business methods remain an excluded category under Section 3(k), and the presence of technical implementation or software execution cannot, by itself, overcome this statutory bar.
Annexure I of the CRI Guidelines 2025 provides a set of illustrative examples that are particularly instructive in the context of AI-based inventions by demonstrating how algorithm-driven techniques may qualify for patent protection when they form part of a larger technical solution. For example, in cases relating to adaptive noise cancellation and medical image enhancement, machine learning models are employed to analyse input signals or image data and dynamically adjust processing parameters, resulting in improved signal clarity, reduced noise, or enhanced diagnostic accuracy. Similarly, examples involving autonomous drone navigation and real-time vehicle stability control illustrate the use of AI-based decision-making models to process sensor data, predict environmental conditions, and generate control outputs that directly influence the operation of physical systems in real time. In each of these scenarios, the algorithmic or mathematical components operate as intermediate mechanisms within an integrated technical system, and patentability is assessed based on the demonstrable technical effects such as improved system performance, responsiveness, safety, or reliability achieved through this integration. Collectively, these examples closely reflect real-world AI deployments and offer practical guidance to applicants and examiners on how AI-driven inventions can be framed and evaluated as technical solutions rather than abstract computational constructs.
Judicial Reinforcement of Substance Over Form
Indian Courts have consistently reinforced that exclusions under Section 3 are to be construed narrowly. In Microsoft Technology Licensing LLC v. Assistant Controller of Patents and Designs (2023), the Court held that an invention should not be rejected as a ‘computer program per se’ solely because it is implemented via software; instead, the patent office must assess whether the claimed invention provides a technical solution to a technical problem and yields a demonstrable improvement in system functioning or efficacy. The Court emphasised that merely executing algorithms on a general-purpose computer is not dispositive of non-patentability if the invention, taken as a whole, produces a tangible technical benefit beyond abstract code. This approach aligns with judicial emphasis on substantive analysis of technical effect over formalistic categorisation and underscores that inventions involving software-based implementations may be patentable when they exhibit concrete technical advantages.
In OpenTV Inc. v. The Controller of Patents and Designs & Anr. (2023), the High Court considered an appeal against the refusal of a patent for a system and method related to interactive media distribution. The Court reaffirmed that Section 3(k)’s exclusion of ‘computer programs per se’ does not categorically bar inventions that exhibit a technical effect or technical contribution; rather, where an invention demonstrates a practical technical improvement, that effect must be considered when assessing patentability. While the Court ultimately upheld the refusal on the particular facts concluding that the claimed subject matter essentially related to a business method, the judgment nonetheless clarified important aspects of the technical-effect analysis under Section 3(k) and referenced the need to distinguish between mere abstract business methods and inventions that deliver tangible technical outcomes in networked or system-level environments. This interpretive emphasis has relevance for AI-based inventions, where innovation often lies in system architecture, intelligent resource management, predictive analysis, or enhanced decision-making mechanisms that produce real-world technical benefits beyond abstract computation.
Synthesising the Legal Position for AI-based Inventions
Under current Indian practice, inventions autonomously generated without identifiable human inventorship are generally not recognised as patentable because AI cannot be recognised as a ‘person’ eligible to be a true and first inventor under Section 6 of the Patents Act. The CRI Guidelines 2025 explicitly distinguish between AI-generated inventions which are not patentable and AI-assisted inventions, which remain patentable provided they are directed by a human inventor and satisfy the standard patentability criteria, including a demonstrable technical effect. Taken together, India’s current patent framework reflects a pragmatic balance. On one hand, statutory exclusions under Section 3(k) remain intact, ensuring that abstract algorithms and business methods are not monopolised. On the other hand, evolving examination practices, refined CRI guidelines, and judicial precedents establish a coherent legal test: AI-based inventions are evaluated on whether they make a technical contribution beyond an abstract algorithm or computational logic. The law now clearly favours an assessment rooted in substance focusing on technical problem to technical solution mapping, system interaction, and real-world technical effects rather than the mere presence of software or AI models.
Applicants are now increasingly guided to draft claims that emphasise system architecture, data processing pipelines, hardware interaction, and measurable technical outcomes rather than abstract model logic. This shift underscores a maturing legal framework that recognises the realities of modern innovation.
Conclusion: Protecting Innovation Through Substance and Technical Contribution
India’s approach to AI-based inventions under patent law is steadily evolving, not through radical legislative overhaul, but through interpretative refinement and procedural modernisation. The judiciary and the patent office have converged on a central principle: innovation deserves protection when it delivers a real technical contribution, regardless of whether it is implemented through AI or traditional technologies.
Looking ahead, continued policy engagement, clearer inventorship norms, and alignment with global best practices may further shape India’s position regarding AI-based inventions. For now, the emphasis remains clear – patentability in India is driven not by labels such as ‘AI’ or ‘software,’ but by the substance of the invention and the technical problem it solves. This innovation-centric outlook is likely to define the next phase of India’s patent jurisprudence in the age of artificial intelligence.
