Generative artificial intelligence can produce text, images, music, software code and other outputs in response to prompts. These capabilities raise questions under copyright, patent and trade mark law. The legal analysis should distinguish at least three matters: the material used to train or operate an AI system, the output generated by the system, and the human or legal person seeking to use or commercially exploit that output.
Copyright and computer-generated works
The Copyright Act, 1957 recognises copyright in original literary, dramatic, musical and artistic works, as well as cinematograph films and sound recordings, subject to the Act. Section 2(d)(vi) identifies, for a literary, dramatic, musical or artistic work that is computer-generated, the person who causes the work to be created as the “author”. This provision is relevant to computer-generated works, but it predates modern generative AI systems and does not expressly resolve every question about prompts, model design, training or the level of human contribution required in a particular case.
The practical inquiry may therefore require close examination of the human contribution and the process by which the output was produced. The selection and arrangement of material, substantive editing, creative decisions, revisions and the extent of independent human expression may be relevant. A prompt alone should not automatically be treated as conferring copyright ownership in every output. Nor should the use of an AI tool automatically be assumed to defeat copyright. The statutory provisions must be applied to the facts and the originality of the work.
Training data and third-party works
The application of these exceptions to large-scale commercial AI training has now received judicial consideration, although the position must be read in light of the procedural stage and facts of the case discussed below.
The fact that a work can be accessed online does not, by itself, mean that it is free for unrestricted copying, training or commercial reuse. The source of the material, the nature and purpose of the use, licences, contractual terms, technological controls and any applicable statutory exception should be examined. AI developers and users should maintain records of dataset provenance, licences and permissions where available.
The 2026 ANI Media v. OpenAI order
A significant development came in ANI Media Pvt. Ltd. v. OpenAI OpCo LLC, I.A. 45300/2024 in CS(COMM) 1028/2024. By an order dated 24 July 2026, the Delhi High Court dismissed ANI’s application for interim injunction. At the prima facie stage, the Court considered the storage and use of ANI’s works for training large language models under Section 52(1)(a)(i) of the Copyright Act, 1957, and found the purpose and fairness factors in favour of OpenAI on the record before it. The Court also considered whether outputs generated through the system substantially reproduced ANI’s works and whether the use caused market substitution or prejudice. The decision was made on an interlocutory application; it did not finally dispose of the underlying suit. The Court expressly confined its observations to the interim application.
The order is important, but it should not be overstated. It concerned an interlocutory application, applied a prima facie assessment to the evidence then available, and expressly stated that its observations would not determine the final outcome of the suit. The principal action therefore remains distinct from the interim ruling. The decision does not create a blanket permission for every AI developer to copy or use every copyrighted work for training. The source and status of the material, the particular training process, the pleaded rights, the evidence of reproduction or memorisation, the market effect and the precise statutory defence remain relevant.
Ownership and contractual allocation
Even where an output is capable of protection, ownership may depend on the circumstances in which it was created and the applicable contractual arrangements. Section 17 addresses first ownership, while Sections 18 and 19 govern assignment and requirements relating to assignments. Employment arrangements, commissioned work, platform terms and client contracts can therefore be significant. Parties should specify who owns protectable human-authored elements, what licences are granted, whether AI tools may be used, and who bears responsibility for third-party rights clearance.
Patent inventorship and AI-assisted inventions
Patent law raises a different set of issues. Section 6 of the Patents Act, 1970 identifies persons entitled to apply, including a person claiming to be the true and first inventor, an assignee of that person, or the legal representative of a deceased person who was entitled to apply. The statutory framework does not expressly set out a process for naming an AI system as an inventor. For an AI-assisted invention, the contribution of the human inventor and the chain of entitlement should therefore be carefully assessed and accurately documented.
Practical safeguards
Organisations using AI tools can reduce uncertainty by maintaining prompt and revision records, documenting human contributions, checking output for similarity to known works and third-party marks, reviewing the platform terms, restricting the upload of confidential or personal data, and recording licences and approvals. For patent-related work, technical notebooks and dated development records may help explain the human inventive contribution. For copyright-sensitive content, a workflow showing drafting, editing and final creative choices can be useful.
AI does not displace existing intellectual property principles. It makes questions of authorship, originality, copying, ownership and responsibility more difficult to identify. Until the statutory framework and judicial treatment develop further, a careful, documented and contractually clear approach is preferable to assuming that every AI output is automatically owned, unrestricted or protectable.