Intellectual Property in the AI Era

Artificial intelligence is generating, assisting, and transforming creative and inventive output at a scale that the existing Indian intellectual property framework was not designed to accommodate. AI tools are now routinely used to generate text, images, code, music, and chemical compound designs; to assist in drafting patent claims and legal documents; and to analyse prior art and identify infringement. The legal questions that arise — who owns AI-generated work, whether AI-assisted inventions are patentable, and what IP rights arise from training data — are actively contested and largely unresolved in Indian law. This article examines where the existing framework applies, where it has gaps, and what organisations and counsel should do now.

Copyright in AI-Generated Works

The Copyright Act, 1957 vests copyright in the “author” of a work — defined as the person who creates it. The Act does not contemplate non-human authors. For computer-generated works, the Act provides that the author shall be “the person who causes the work to be generated.” However, Indian courts have not yet determined whether a work generated by an AI model with minimal human direction qualifies for copyright protection, or in whom copyright vests. Pending judicial clarification, the safest approach is to ensure demonstrable human creative contribution beyond the prompt — the AI functions as a tool, with the human contributing original expression at a level sufficient to sustain authorship. Where AI generation is extensive, the copyright position of the output should be assessed before it is commercially exploited.

AI-Assisted Inventions and Patent Law

The Patents Act, 1970 requires an invention to be made by a “person.” The Indian Patent Office has not yet addressed whether inventions in which an AI system made a substantive inventive contribution are patentable, or who should be named as inventor. In jurisdictions that have considered the question — including the UK, US, and EU — applications naming AI systems as inventors have been rejected. Until India takes a definitive position, organisations should ensure that human inventors are identified who made substantive contributions to each inventive step, and that the role of AI tools is documented in development records without being named on the application. Patent applications for AI-assisted inventions should be filed without delay, as the prior art landscape in AI evolves rapidly and delays compress the novelty window.

Training Data and Third-Party IP Risk

Training AI models on large datasets raises significant IP risk. Where training data includes copyrighted works — text, images, code, or music scraped from public sources — the reproduction and processing of those works may constitute infringement of the reproduction right under Section 14 of the Copyright Act, 1957, unless an exception applies. India’s “fair dealing” provisions under Section 52 are narrowly drafted (research, private study, criticism, review, and news reporting) and are unlikely to cover large-scale commercial training data processing. Organisations developing or procuring AI models should conduct training data IP audits and seek representations and warranties from AI vendors about the provenance and licensing of training data, with appropriate indemnity provisions for third-party IP claims.

Licensing, Contracts and Enforcement

Where an organisation uses AI to generate deliverables for clients — documents, code, reports, or designs — the contractual arrangements with both the AI provider and the client must address IP ownership clearly. Standard AI platform terms typically grant the user a licence to the output rather than assigning copyright, and many platforms disclaim any warranty that the output does not infringe third-party IP. Organisations should review their AI vendor agreements for IP assignment and indemnity clauses, and include appropriate representations in client-facing agreements addressing the AI-assisted nature of the deliverable. The IT (Intermediary Guidelines and Digital Media Ethics Code) Rules, 2021 also prohibit the generation and publication of deepfakes and synthetic content impersonating real persons, creating an additional enforcement exposure for organisations deploying AI in content-intensive workflows.

Trade Secrets and AI Model Protection

AI models and training datasets represent significant proprietary value that falls outside the scope of registered IP rights in most cases. Protection depends on trade secret law — the model weights, training data, fine-tuning methodology, and system architecture are confidential information that must be protected through appropriate contractual and access-control measures. Employment agreements and contractor agreements should include clear IP assignment clauses and confidentiality obligations covering AI-related work product. Access to model infrastructure should be restricted to those with a business need, and departing employees with access to AI development environments should be subject to additional offboarding procedures including forensic verification of data return and deletion.

Regulatory Developments and Compliance Horizon

India does not yet have an AI-specific regulatory framework, but several regulatory bodies are developing their approach. The Ministry of Electronics and Information Technology has published advisory frameworks on responsible AI use, and sector-specific regulators — SEBI, RBI, IRDAI — have begun addressing AI use in financial services and insurance. The EU AI Act, enacted in 2024, has extraterritorial effect on high-risk AI systems deployed to EU users, which may affect Indian AI companies with European operations or customers. Businesses building or deploying AI systems should monitor these developments and engage legal counsel on IP structuring, data licensing, and regulatory compliance as the framework matures — proactive structuring now avoids costly retrofitting when mandatory requirements take effect.