Software and AI Patents in India: What’s Allowed and What’s Not (2025 Update)

The digital age has made software and Artificial Intelligence (AI) the new frontier of innovation. However, patenting these technologies in India has long been a complex and challenging endeavor. The legal framework, primarily centered on Section 3(k) of the Patents Act, 1970, has created a narrow path for innovators. As of 2025, a series of judicial precedents and the landmark issuance of the new Revised Guidelines for Examination of Computer Related Inventions (CRIs) by the Controller General of Patents, Designs & Trade Marks (CGPDTM) have brought much-needed clarity.

This article dissects the current landscape of software and AI patentability in India, outlining what is permissible and what is not, based on the latest guidelines and prevailing legal interpretations.

The Foundation: Section 3(k) of the Patents Act, 1970

At the heart of India's patent regime for software and AI is Section 3 of the Patents Act, which provides a list of non-patentable inventions. Specifically, Section 3(k) excludes:

“a mathematical or business method or a computer programme per se or algorithms”

The key term here is "per se," which has been the subject of extensive legal interpretation. The legislative intent behind its inclusion was to prevent the monopolization of abstract ideas and algorithms without any tangible, real-world technical application. The challenge for patent applicants has always been to demonstrate that their invention goes beyond being a "computer programme per se."

The Evolving Standard: The "Technical Effect" Test

To navigate the strictures of Section 3(k), Indian courts and the IPO have developed and consistently applied the "technical effect" test. This principle was solidified by landmark rulings, most notably the Delhi High Court's decision in Ferid Allani v. Union of India and reaffirmed in cases like Microsoft Technology Licensing LLC v. Assistant Controller of Patents.

The core of this test is simple: an invention is patentable if it solves a technical problem and produces a "technical effect" or "technical contribution," even if it is implemented using software. The invention must demonstrate a tangible improvement in the functioning of the computer system or a related technical field, not just a new method of computation or business transaction.

What’s NOT Allowed: The Excluded Subject Matter

Based on the legal framework and the 2025 CRI Guidelines, the following are generally excluded from patentability:

1. A "Computer Programme per se": This refers to the source code or a computer program in its abstract form, without being tied to a specific hardware or technical problem. A mere set of instructions or a flowchart is not patentable.

2. Algorithms: A pure algorithm or a mathematical formula is not patentable. This exclusion is intended to ensure that fundamental scientific principles and mathematical discoveries remain in the public domain. For instance, a new sorting algorithm or a cryptographic method, in itself, is not patentable.

3. Business Methods: A method of doing business, even if it is implemented on a computer system, is not patentable. The IPO will look at the substance of the invention. If the core of the invention is a business process, such as a new e-commerce model or a financial trading strategy, it will be rejected under Section 3(k).

4. Presentation of Information: This exclusion, under Section 3(n), is often cited in conjunction with Section 3(k). It prohibits the patenting of a new way of presenting information, such as a new user interface layout or a new data visualization method, if it lacks a demonstrable technical effect.

What IS Allowed: The Patentable Path for Software and AI

The new 2025 CRI Guidelines, the result of a multi-stage consultation process, have created a more structured and predictable framework for patenting AI and software. An invention is likely to be patentable if it demonstrates a technical contribution in one of the following ways:

• Improved Hardware Functionality: The software or AI model enhances the efficiency, speed, or performance of the hardware it runs on. For example, an algorithm that reduces a computer's memory usage or a program that optimizes data processing for a specific hardware component would be considered patentable.

• A Technical Solution to a Technical Problem: The invention uses software to solve a genuine technical problem. The 2025 Guidelines provide specific examples for AI/ML inventions, such as:

  • An AI model for improving network security by detecting and mitigating cyber threats.

  • A machine learning algorithm for optimizing data compression techniques, leading to more efficient data storage and transmission.

  • An AI system for enhancing the robustness of a network against packet loss.

  • The use of AI for improving the performance of a robotic arm in a manufacturing process.

• System and Method Claims: The 2025 Guidelines clarify that both system and method claims are permissible, provided they are supported by the patent specification. This means a patent can be granted for a system (e.g., "a system comprising a processor and a memory configured to…") or a method (e.g., "a method for..."). The claims must, however, be tied to the hardware and demonstrate a technical effect.

Critical Considerations for AI Patent Applications

The 2025 Guidelines have a dedicated section for AI, ML, and DL, with a strong emphasis on sufficiency of disclosure. To successfully patent an AI-based invention, applicants are now required to provide precise details, including:

• The nature and handling of training data: The guidelines stress the need to describe how the AI model was trained.

• Algorithmic procedures: Applicants must provide sufficient detail on the activation functions, learning strategies, and other technical aspects of the algorithm.

• Environment dynamics: For inventions using reinforcement learning, the disclosure must explain the environment in which the AI agent interacts.

The message is clear: the IPO will be scrutinizing AI patent applications to ensure they are not merely abstract algorithms but provide a complete, well-disclosed technical solution.

Conclusion

The patentability of software and AI in India in 2025 is no longer a matter of simple black and white. While the statutory exclusion under Section 3(k) remains, the judicial interpretations and the new CRI Guidelines have created a more nuanced and encouraging environment for innovators. The key takeaway is to shift the focus from the software itself to the technical problem it solves and the technical effect it produces.

For startups, developers, and researchers in the AI space, the path to a patent grant in India now requires a strategic approach. The patent application must be meticulously drafted, clearly articulating how the invention provides a tangible technical advancement over the prior art. By emphasizing the integration of software with hardware, and by clearly demonstrating a real-world technical contribution, innovators can successfully secure patent protection for their cutting-edge creations in India's rapidly evolving intellectual property landscape.