Beyond the Source Code: Can India’s Patent Framework Accommodate Black-box Software?

Introduction

The rapid growth of Artificial Intelligence (AI), machine learning, and data-driven technologies has increased the significance of “black box” software systems. A black box algorithm is generally understood as a computational model where the inputs and outputs are known, but the internal decision-making process is either opaque or not readily understandable. While such technologies have significant commercial value, they raise difficult questions under patent law because patents are granted in exchange for public disclosure of the invention.

In India, the patentability of black box software is governed principally by Section 3(k) and Section 10 of the Patents Act, 1970. Section 3(k) excludes “a mathematical or business method or a computer programme per se or algorithms” from patentability, while Section 10 mandates sufficient disclosure of the invention. Indian courts, particularly in Ferid Allani v. Union of India [1], Microsoft Technology Licensing LLC v. Assistant Controller of Patents and Designs [2] and OpenTV Inc. v. Controller of Patents and Designs [3] have clarified that software-based inventions are not automatically excluded from patent protection and must be assessed on the basis of their technical contribution and technical effect. However, satisfying Section 3(k) is only the first hurdle; the more persistent difficulty for black-box AI lies in Section 10 disclosure mandate, which this article examines in greater depth below.

How Does a Black Box Algorithm Work?

A black box algorithm functions by receiving input data, processing it through a computational model, and producing a result. In AI and machine-learning systems this may involve numerous variables, training parameters, neural-network layers, or statistical models, so a user can verify the output without explaining the precise pathway through which it was generated. Patent law, however, does not protect mere results as it protects a technical solution to a technical problem. An applicant must explain the invention through its architecture, processing steps, and technical implementation; a bare assertion that an algorithm receives information and generates a prediction will not suffice.

Patentability of Black Box Software under Section 3(k)

One of the primary challenges for black box software inventions is the exclusion contained in Section 3(k) of the Patents Act. Examiners often object that software-based inventions are merely algorithms or computer programs and therefore not patentable.

However, Indian jurisprudence has progressively adopted a technology-oriented approach. The most important authority is Ferid Allani (supra), where the Delhi High Court held that computer-related inventions should not be rejected merely because they involve software. The Court observed that the expression “computer programme per se” was deliberately incorporated to ensure that genuine technological inventions are not excluded from patent protection. The correct inquiry is whether the invention demonstrates a technical effect or technical contribution i.e. for an instance, improved computer functioning, network efficiency, cybersecurity, processing speed, or resource consumption.

Therefore, a black box algorithm is not automatically unpatentable. What matters is whether the invention produces a measurable technical effect beyond a mere abstract algorithm.

Compliance with Section 10: The Real Bottleneck

Even if an AI-based invention overcomes the exclusion under Section 3(k), it must also comply with the disclosure requirements of Section 10(4) of the Patents Act, which require the complete specification to fully describe the invention, disclose its operation, set out the best method known to the applicant for performing it, and conclude with claims defining the scope of protection. Nevertheless, patent law requires sufficient technical disclosure to enable a person skilled in the art to perform the invention, even though disclosure of the source code is not mandatory. Accordingly, the specification should adequately explain the system architecture, operational workflow, implementation steps, technical features, and the technical effect achieved. In this sense, Section 10(4)’s ‘person skilled in the art’ test does the same job as the enablement/PHOSITA standards used elsewhere: it asks whether a skilled worker, not just the inventor, could actually carry out the invention from what’s disclosed. Judicial decisions have reinforced this requirement. In Ferid Allani (supra), the Delhi High Court stressed the need to identify the invention’s technical contribution, which is only possible through adequate disclosure. Similarly, in Microsoft (supra) and OpenTV (supra), the Court emphasized that patentability must be assessed based on the technical advancement and technical solution disclosed in the specification. Further, the “best method” requirement under Section 10(4)(b) obliges applicants to disclose the preferred implementation of the invention, including the relevant AI architecture or system configuration where essential, without requiring disclosure of every line of source code.

Examination Objections Commonly Raised by the Controller

Three objections recur in the prosecution of black-box AI patent applications. Firstly, under Section 3(k), the Controller may contend the claim is merely an algorithm or computer program per se; applicants may respond with Ferid Allani to demonstrate a technical effect such as improved processing efficiency or cybersecurity, with Microsoft for the principle that an invention must be assessed as a whole rather than by isolating its software components, and with OpenTV for the proposition that the relevant inquiry is a technical solution to a technical problem, irrespective of software implementation. Secondly, under Section 10(4), the Controller may object that the specification describes only the desired outcome without disclosing the underlying mechanism; this may be met with system architecture, workflow diagrams, functional algorithms, implementation examples, and preferred embodiments sufficient to enable a skilled person, without revealing proprietary source code. Lastly, the Controller may argue the invention merely automates a known process without inventive step; the applicant must then show novelty and a demonstrable technical contribution over the prior art, again anchored in the same three authorities.

Conclusion

Black box software presents a complex challenge under Indian patent law, as internal functioning may not be readily explainable. Such inventions are not excluded merely for involving software or algorithms: Ferid Allani, Microsoft, and OpenTV confirm that software-related inventions may be patented where they show a technical effect, contribution, or solution. But that technical-effect inquiry is only the first hurdle. Applicants must also satisfy Section 10(4)’s enabling-disclosure standard, and black-box AI strains it because the very opacity that gives the technology commercial value can be what prevents a skilled person from reproducing the claimed result. Until the Patent Office issues AI-specific examination guidelines, on training-data disclosure, architecture-level description, and a workable reproducibility threshold i.e. applicants are well advised to anchor specifications in disclosed architecture, workflow, and implementation detail, sufficient to withstand both a Section 3(k) and a Section 10(4) challenge. Where these requirements are met, black box software inventions have a strong basis for protection under Indian law; where they are not, the gap between commercial opacity and legal disclosure remains a growing risk.