<\/p>\n
India has introduced updated guidelines in 2025 to streamline the legal framework for AI-related assets, differentiating between human-involved and autonomous AI inventions across patents, copyright, trade marks, and designs, while highlighting ongoing challenges around training data and trade secrets.<\/p>\n<\/div>\n
India\u2019s intellectual property regime now has a clearer, if still imperfect, framework for artificial intelligence. The country continues to protect AI-linked assets through the four familiar channels of patents, copyright, trade marks and designs, but the practical test is no longer whether an AI system is involved; it is which part of the asset the law can actually recognise, and on whose behalf. The latest shift came with the Revised Guidelines for Examination of Computer Related Inventions, issued in July 2025, which the government says were designed to improve consistency in examining AI, machine learning, deep learning, cloud computing, quantum computing and blockchain-related inventions.<\/p>\n
For patents, the central issue remains Section 3(k) of the Patents Act, which excludes mathematical methods, business methods, computer programs per se and algorithms. The revised CRI Guidelines 2025, together with recent High Court rulings, emphasise that applicants must show a genuine technical contribution or technical effect, not just automation dressed up as innovation. The Indian Patent Office has also sharpened the disclosure burden for AI and ML inventions, expecting applicants to explain model structure, data characteristics, training approach and validation in enough detail to satisfy sufficiency of disclosure.<\/p>\n
The inventorship point is now firmly settled in India\u2019s administrative guidance: AI cannot be listed as the inventor. The CRI Guidelines treat AI as a tool, not a legal person, and keep the application route tied to a human applicant under Section 6. That makes a distinction between AI-assisted inventions, which may still be patentable if they meet the ordinary tests, and AI-generated inventions that are produced with little or no human intervention, which the framework treats very differently.<\/p>\n
Copyright follows a different logic. Indian law already contains a computer-generated works provision, and the current position is that the author of such a work is the human who causes it to be created. That leaves room for protection where a person meaningfully directs, curates or shapes the output, but it does not automatically extend authorship to the machine itself. The practical result is that ownership questions will often turn on contracts, prompt design, editorial control and the surrounding chain of human contribution.<\/p>\n
The unresolved pressure point is training data. India still has no dedicated text-and-data-mining exception, and the Copyright Act\u2019s fair-dealing provisions were not drafted for large-scale model training. That is why the pending litigation brought by ANI Media against OpenAI matters so much: it is expected to become the first major judicial test of whether storing copyrighted material for training, or generating outputs from such material, can be justified under existing Indian law.<\/p>\n
Trade marks are less legally dramatic but no less important commercially. The application is still filed by a human or corporate applicant, and registrability depends on the ordinary tests of distinctiveness and confusion. AI may now be used to speed up clearance searches, monitoring and portfolio management, but it does not change the basic filing position. The real question for AI-generated branding is whether the mark functions as a source identifier in the mind of the consumer.<\/p>\n
Design law raises its own authorship issue. A visual interface, product shell or other ornamental feature may qualify for registration, but the law still assumes an identifiable human author. If the design is merely the output of an autonomous system with no meaningful human direction, its status becomes far more vulnerable. In practice, applicants will need records showing human selection, iteration and curation if they want an AI-assisted design to stand up at the registration stage.<\/p>\n
Where India is still weak is in confidential know-how. There is no standalone trade-secret statute, so training data, model weights, prompts and fine-tuning methods depend on contract, confidentiality obligations, employment assignments and access controls. For AI businesses, that means the legal protection stack is mixed: patents for technical inventions, copyright for human-directed expressive works, trade marks for brands, designs for appearance, and contracts for the data and model assets that sit outside all four regimes.<\/p>\n
Comparatively, India now sits in line with other major jurisdictions on one important point: AI is not treated as an inventor. The United States, the United Kingdom and the European Patent Office have all taken the same broad position on inventorship, though they diverge on training-data exceptions and trade-secret architecture. The Indian position is therefore no longer unusual on invention ownership, but it remains incomplete on data use, where the law has yet to catch up with how modern AI systems are built and trained.<\/p>\n
Inspired by headline at:<\/strong> [1]<\/a><\/sup><\/p>\n