Current Indian Law

The Copyright Act of 1957 is a great legislation as it expressly defines the author for computer-generated content in case of literary, dramatic, musical or artistic work as "the person who causes the work to be created".

 The words used by lawmakers on literal interpretation put humans into authorship and allow  Indian courts to attribute authorship to a human connected to any output generated by AI. But it does not specify which human would be given authorship: Would it be the person who gives the prompt to the AI, or the developer of the AI, or the platform operator? It leads to 2 consequences:

1.     The copyright act clearly denies recognising a machine as an author: When content is generated through AI, the AI cannot be considered as an author, as Indian laws do not give machines the status of a legal person,

2.      Practical disputes arise about which human “caused” the work, and that is a fact-sensitive question.

Comparative Approaches to Human Authorship and AI-Generated Content in U.S. and Indian Law

The U.S. Copyright Office’s guidance requires human authorship for registration and instructs examiners to identify and exclude material generated solely by AI from protection. Recent U.S. appellate decisions have reaffirmed that works produced without human creative (often referred as sweat of the brow doctrine) contribution do not qualify for copyright.] At the same time, courts have allowed protection for works where humans make meaningful creative contributions to AI-produced content. 

In India, public discourse and litigation (including disputes involving large platforms and AI training data) are already testing how doctrines such as “derivative works” and fair dealing apply to model training and output.

Key policy trade-offs

1.Incentives vs. Access. 

Granting broad copyright in AI outputs would create new exclusive rights that could reward prompt engineers, platform owners or model trainers, but may also raise barriers to access and accelerate copyright proliferation, hindering future creativity. 

2. Identification and enforcement costs. 

Determining who “caused” a work can be complex and expensive: was the output essentially the model’s autonomous generation, the user’s prompt, or the developer’s architecture and dataset curation? Misallocation risks produce litigation and uncertainty. 

3. Dataset fairness and liability.

The AI Models are usually criticised for being trained on different third-party copyrighted works, and that too without the consent of the copyright owners.

4. Public order and personality rights. 

Along with copyright, Indian courts are trying to reduce use of celebrity images and voice clones in order to protect personality rights and to maintain public trust. The courts want platforms to take more responsibility and accountability to ensure that AI-generated content does not create a hindrance in society.

Practical approach for India

1. Redefining "Causes" in Human Authorship

Current laws favour whoever "causes" a work to be created as per S.2(d)(vi) of copyright act, but that is too vague and no longer clear in the age of generative AI. Rather than relying on vague interpretations, we need to move beyond them by:

Defining the Actor: The Copyright Office should issue clear rules to distinguish between the prompt submitter, the Developer of AI, and the data feeder.

Doctrine of Sweat of Brow: Rights should not be given automatically. A person must prove that the output was due to a specific input and involved some effort, and that it was not just chance. By ensuring that human creativity remains central and a major basis for granting copyright protection.

2. Avoiding "Copyright Maximalism"

Copyright laws are a wonderful tool, but they should not protect every AI creation.

A Multi-Front Defence: Issues such as deepfakes and voice cloning must be addressed through personality rights and intermediary liability rules, rather than relying solely on copyright.

Sui Generis Solutions: Instead of stretching copyright laws. Indian laws must adopt sui generis rights to ensure that artists displaced by AI training are compensated using industry-wide funds, similar to Spotify's practice. It pays a collective pool to the industry, which is then divided among the artists based on how much people actually listen to them. This seeks to bring that same aspect to the world of AI training data.

Conclusion — balancing innovation, creators and the public

Even though the current "computer, generated" clause in the Act offers a workable backing still it safeguards the people who have a considerable input in the creation of AI outputs, and at the same time, it imposes transparency, source identification, and fair, use oriented limitations as remedies to both the situation of a single player monopolizing the public domain and the unremunerated use of creators' works in the training process.

Additional regulatory tools, for example, dataset licensing, disclosure regimes, personality protections, and sectoral licensing schemes, will be necessary to establish a fair, innovative, and friendly balance.