The Patentability of Artificial Neural Networks: Lessons from the UK Supreme Court for Indian Patent Law

Introduction

Can an Artificial Neural Network (“ANN”) be patented? This question has puzzled courts across various jurisdictions, and the recent UK Supreme Court judgment in Emotional Perception AI Ltd. v. Comptroller General of Patents, Designs and Trade Marks contains some valuable lessons for Indian patent law. Although UK and Indian patent law vary significantly, the UKSC offers a structured methodology for examining Computer-Related Inventions (“CRI”), providing a valuable blueprint that India can incorporate into its evolving jurisprudence on AI patentability.

The UKSC’S Blueprint

The UKSC judgment, delivered on 11 February 2026, fundamentally reshapes the UKs approach to patenting ANNs. The Court definitively held that an ANN is a “program for a computer” within the meaning of article 52(2)(c) of the European Patent Convention. The Court reasoned that “the ANN constitutes, in essence, a set of instructions to manipulate data in a particular way so as to produce a desired result. In other words, an ANN is a program for a computer.”

However, the real significance lies in the Court’s rejection of the old Aerotel “technical contribution” approach. That earlier methodology, endorsed by the Court of Appeal in Aerotel Ltd. v. Telco Holdings Ltd., had asked examiners to identify “what the inventor has really added to human knowledge” and then determine whether that contribution fell solely within excluded subject matter. In practice, this meant leapfrogging over the threshold question of whether an “invention” existed at all, and instead conflating the inquiry with considerations of novelty and inventive step.

In place of this methodology, the UKSC endorsed a three-step analysis derived from the EPO Enlarged Board’s decision in G1/19 (Bentley Systems/Pedestrian Simulation). Step one asks whether the claimed subject matter constitutes an “invention” within the meaning of Article 52 of the EPC. This is a deliberately low threshold. Any claim involving technical means, including a general-purpose computer programmed to perform a method, satisfies this requirement.

Step two is what the Court called “the intermediate step”. It requires a careful examination of the claim to identify which features contribute to the overall technical character of the claim, and which are non-technical features “as such”. Features that would be excluded if viewed in isolation (such as mathematical methods, business schemes, and abstract algorithms) may be taken into account for inventive steps if they interact with the technical features to give the claim an overall technical character.

Step three restricts the assessment of novelty and inventive step to those features that survive the intermediate filter. Only features contributing to the technical character of the invention may support patentability. Non-technical features, even if novel and non-obvious when viewed in isolation, are excluded from consideration. This ensures that patents are granted only for genuine technical contributions, not for abstract ideas merely implemented on a computer.

The Indian Paradigm: Decoding Section 3(K)

Section 3(k) of the Patents Act, 1970, presents a more complex landscape. It excludes from patentability “a mathematical or business method or a computer programme per se or algorithms”. Unlike UK and European law, where “as such” qualifies all excluded categories, India’s “per se” attaches only to “computer programme”. The 2025 Computer-Related Inventions Guidelines elaborate on this distinction.

The 2025 CRI Guidelines provide a structured approach for identifying each exclusion criterion. For mathematical methods, paragraph 4.5.1.1 asks whether the solution lies in abstract mathematical processing or is part of a larger technical process. For algorithms, paragraph 4.5.3.1 differentiates between abstract sequences of steps and enabled implementations of a technical nature. For computer programmes per se, paragraph 4.5.4.1 mandates construing the substance of the claim to determine whether the identified technicality results in a larger technical effect, beyond a merely incidental one.

The Critical Divergence

If the UKSC’s first-step “any hardware” test were mechanically applied in India, it would be wholly insufficient. An ANN implemented on a server rack clearly involves hardware, but under Section 3(k), that is merely the starting point. The claim could simultaneously attract multiple exclusions: as a mathematical method (involving backpropagation calculations), as an algorithm (a set of rules for processing data), and as a computer programme per se (if it merely automates semantic comparison without technical effect).

Yet discarding the UKSC judgment as irrelevant to India would be a profound error. The true lesson lies not in the outcome but in the process of analysis.

The lesson: Importing the “Intermediate Step”

The UKSC’s intermediate step offers India a sophisticated analytical tool that aligns remarkably well with the 2025 CRI Guidelines. Paragraph 4.5.4.1 of the Guidelines already requires “construing the substance of claimed invention as a whole and identifying the essential technical features”. This mirrors the UKSC’s mandate to identify which features contribute to technical character.

India should consider formalising a structured approach adapted from the UKSC’s framework:

Step 1: Construct the Claim. Identify the invention as claimed, consistent with established practice.

Step 2: Map Features to Section 3(k) Exclusions. Using the 2025 CRI Guidelines, systematically identify which claim elements fall within the four exclusions: mathematical method, business method, computer programme per se, or algorithm. This requires substance-over-form analysis.

Step 3: The Indian Intermediate Step. Filter out features that are excluded “as such”, namely, mathematical methods without technical application, algorithms devoid of implementation specifics and computer programmes that produce no technical effect beyond mere automation. Crucially, identify which features, even if falling within excluded categories, “interact with the technical subject matter of the claim” to contribute to technical character. This mirrors the UKSC’s inquiry: “Whether such features contribute to the technical character of the invention has to be assessed in the context of the invention as a whole”.

Step 4: Assess Inventive Step. Restrict the inventive step analysis to features surviving the filter. The inventive step must reside in the technical features, i.e. how the ANN is implemented, its specific architecture, its integration with hardware, or the resulting technical effect (e.g., improved processor efficiency, reduced latency, enhanced control of a physical device).

Practical Application To ANN Inventions

Applying this structured approach to an ANN-based recommendation system, such as the one in Emotional Perception, leads to a more rigorous analysis. In such a system, the mathematical methods of backpropagation and weight adjustment, when viewed alone, might fall in the mathematical method exclusion, and the algorithmic sequences of processing data through layers might lead to the algorithm exclusion. However, if the claim properly explains how the ANN interacts with a database, a communications network and a user device to produce better recommendations through technical improvements in data processing, these features will be viewed as a whole and may survive the filter, thus falling within the “inventive step” criterion.

This approach prevents the outright rejection of AI-based inventions while also ensuring that only genuine technical contributions are patented.

Recommendations for Indian Patent Law

Firstly, Indian courts and the Patent Office should adopt a structured approach to the patenting of AI-based inventions (as described earlier), incorporating the intermediate step into the Section 3(k) analysis and drawing on the UKSC’s reasoning. The 2025 CRI Guidelines provide fertile ground for the development and adoption of such an approach.

Secondly, the “substance-over-form” approach must be implemented to properly differentiate excluded matter from actual technical features. Claims presenting abstract ideas as technical implementations must be identified and filtered appropriately.

Thirdly, India should resist the temptation to mechanically follow foreign outcomes while ignoring foreign methodologies. The UKSC’s judgment is valuable not because it answers whether ANNs are patentable, but because it provides a sophisticated analytical framework for asking the right questions.

Conclusion

The UKSC has given the world an incredible approach for analysing AI-based inventions. By incorporating the “intermediate step” approach into Section 3(k) of the Indian Patents Act, 1970, we can develop a more rigorous and transparent system for examining AI-based inventions.

However, the UKSC has only provided a blueprint, and now, Indian law must build upon it.

Author: Aarush Mishra, in case of any queries please contact/write back to us via email to [email protected] or at IIPRD. 

References

• Emotional Perception AI Ltd v Comptroller-General of Patents, Designs and Trade Marks [2026] UKSC 3. <https://www.supremecourt.uk/cases/uksc-2024-0131>.
• Patents Act 1970. <https://www.indiacode.nic.in/handle/123456789/1392?view_type=search>.
• Office of the Controller General of Patents, Designs and Trade Marks, Guidelines for Examination of Computer Related Inventions (CRIs) 2025 <https://ipindia.gov.in/writereaddata/images/pdf/CRI/1.%20GUIDELINES%20FOR%20EXAMINATION%20OF%20COMPUTER%20RELATED%20INVENTIONS%20(CRIs)-%202025.pdf> accessed 7 March 2026.
• Convention on the Grant of European Patents (European Patent Convention) 1973. <https://www.epo.org/en/legal/epc/2020/a52.html>.
• Patents Act 1977. < https://www.legislation.gov.uk/ukpga/1977/37>.
• Mishcon de Reya LLP, ‘Patentability of Artificial Neural Networks: Revolution or Evolution?’ (Mishcon de Reya, 2026) <https://www.mishcon.com/news/patentability-of-artificial-neural-networks-revolution-or-evolution> accessed 7 March 2026.
• Bird & Bird LLP, ‘Patentability of Artificial Neural Networks to Be Decided by UK Supreme Court on Wednesday 11 February’ (Bird & Bird, 2026) <https://www.twobirds.com/en/insights/2026/uk/patentability-of-artificial-neural-networks-to-be-decided-by-uk-supreme-court-on-wednesday-11-februa> accessed 7 March 2026.