Most people who deal with Indian patent practice know that software is difficult here, and stop there. The received position is that India does not grant software patents, and applicants with software-implemented inventions either avoid the jurisdiction or file and hope.

The received position is wrong, and it has been wrong for some time. What section 3(k) excludes is a "computer programme per se" — not software, not computer-implemented inventions, and not claims that happen to involve code. Those two Latin words have been doing an extraordinary amount of work since 2002, and the question of what they leave available has been answered inconsistently by examiners for most of that period.

On 29 July 2025 the Patent Office published revised guidelines for examining computer-related inventions, superseding the 2017 version. They are the most detailed statement of the Office's position that has been available, and they matter more than guidelines usually do — because in this area the gap between the statute and the examination practice has been the applicant's central problem.

What Section 3(k) Actually Says

The provision excludes four things: a mathematical method, a business method, a computer programme per se, and an algorithm.

Three of those are absolute. The fourth carries a qualifier that has generated fifteen years of litigation. A computer programme per se is excluded; a computer programme that is not merely a programme in itself — that produces something beyond the running of the code — is not caught by the words Parliament used.

The Delhi High Court has been the main engine of this reading. Ferid Allani established that a computer-implemented invention demonstrating a technical effect or technical contribution falls outside the exclusion. Microsoft Technology Licensing pushed further into how that assessment should be conducted. The 2025 guidelines draw on nineteen decisions in total, which tells you how much of this framework has been built by courts correcting the Office rather than by the Office itself.

The Test the Guidelines Adopt

The operative question is whether the claimed invention involves a technical effect or technical contribution.

That formulation is not new, but its consolidation into the examination guidance is. An invention that improves how a machine functions — faster processing, reduced memory use, better security, more efficient resource allocation, an improvement in the operation of hardware or of a technical process — contributes something technical. An invention whose entire contribution is a smarter business rule, or a mathematical relationship, or an abstract sequence of steps expressed in code, does not, however sophisticated the code.

The guidelines also confirm that both system and method claims are permissible in independent form, provided the specification supports them. That sounds procedural and is not. Claim-format objections have been a routine feature of Indian CRI prosecution, with applicants pushed towards system claims on the assumption that method claims are inherently vulnerable under section 3(k). Confirming both are available removes a constraint that has distorted drafting for years.

Annexure I carries a non-exhaustive set of worked examples of allowable and non-allowable subject matter under each limb of the exclusion. For prosecution purposes that annexure may be the most useful part of the document, because it shows the Office's reasoning applied rather than stated.

Where the Emerging Technologies Sit

The 2017 guidelines predate the deployment of machine learning at commercial scale, and it showed. The 2025 version takes a technology-specific, example-driven approach to artificial intelligence, machine learning, deep learning, blockchain and quantum computing, setting out scenarios and the disclosure each requires.

The disclosure point is the one applicants underestimate. An AI-related application that describes a model architecture in general terms and asserts an improvement has a weak position on both technical effect and sufficiency. What is required is the specific technical problem, the specific mechanism by which the claimed approach addresses it, and enough detail for a skilled person to work the invention. A model that performs better is not, by itself, a technical contribution — the specification has to establish what changed technically to make it perform better.

What This Does Not Change

Guidelines bind examiners in practice and not in law. They do not amend section 3(k), and they do not bind the courts. An applicant refused despite meeting the guidance still has section 117A, and a patent granted in reliance on the guidance can still be opposed and revoked.

More practically, guidance does not immediately change habits. The gap between what the Delhi High Court has said about technical effect and what applicants encounter in examination reports has persisted for years despite the case law being clear. Whether the 2025 document narrows that gap depends on adoption at the examination stage rather than on the quality of the drafting — and that will not be visible for another year or two.

The Takeaway

For software and AI-driven businesses that have written India off, the position deserves revisiting. The exclusion is narrower than its reputation, the framework for engaging with it is now documented, and the worked examples give a drafting target that did not previously exist in accessible form.

The work has to be done in the specification. Technical effect is established by what the document discloses about the technical problem and the technical means, and no amount of argument at the examination stage rescues a specification drafted as though the invention were a business idea implemented on a computer. Applications drafted for jurisdictions with different exclusions frequently arrive in India without that material, and it cannot be added later.

For anyone already in prosecution and holding a section 3(k) objection: the objection is answerable, the answer is a technical-effect argument grounded in the specification and supported by the case law the Office now cites, and the annexure of examples is the reference against which the response should be built.