Generative AI — the models that write text, compose music, create images and even draft films — has exploded into everyday creative workflows. For law students in India the central question is deceptively simple: when an AI produces a work, who (if anyone) owns the copyright?

The answer sits at the intersection of statutory text, administrative guidance abroad, and rapidly developing case law — and today the law still tends to favour human authorship. Below I explain the statutory starting point in India, relevant international developments, judicial trends, practical work-arounds (contracts/licences) and what to watch for next.

1. The statutory starting point in India

The Indian Copyright Act, 1957 sets the baseline. Two provisions matter most:

• Section 2(d) defines “author” and — crucially for AI — explicitly treats computer-generated works differently: for “any literary, dramatic, musical or artistic work which is computer-generated, the person who causes the work to be created” is treated as the author. That wording presumes a human cause and links authorship to the person who initiates/controls the process.
• Section 13 & Section 17: Section 13 lists the classes of protectable works; Section 17 provides that copyright initially vests in the person who creates the work (the author) — again pointing to a human origin. Registration is evidence but not a precondition of ownership.

The net result under current Indian statute: copyright protection is most straightforward where a human author can be identified — either because a person wrote the content or because someone “caused” a computer-generated work to be produced (for instance by setting the parameters, curating inputs, or issuing specific instructions). If a model runs fully autonomously and no human can be said to have caused the work, Indian statutory language leaves a grey zone.

2. International guidance and judicial trends

Abroad, agencies and courts have recently emphasised human authorship. The U.S. Copyright Office’s multi-part AI study and guidance (Part 2 on copyrightability) explains that copyright protects human authorship and that applicants must identify their human contributions when registering works that contain AI-generated material.

Courts have followed suit. In Thaler v. Perlmutter and related proceedings, litigation over whether an artwork generated by an AI (DABUS) could be registered drew a firm statement that human authorship remains a bedrock requirement — registration was denied where the applicant sought protection for an autonomously generated work. Recent appellate decisions in the U.S. have upheld that position. These developments are persuasive globally and are already shaping policy debates in India.

3. Indian judicial and regulatory movement

India has not yet produced a definitive Supreme Court ruling about purely AI-generated copyright. However, courts are already confronting AI issues in related areas — for example, the Bombay High Court recently ordered removal of highly realistic deepfakes that used a celebrity’s likeness, stressing harms from AI-generated content and protecting personality/right of publicity considerations. Such orders indicate courts will use existing legal doctrines (defamation, personality rights, injunctions) to curb abusive AI uses even before specific copyright answers arrive.

4. Practical legal consequences and doctrinal implications

• Who can register? When a human has a demonstrable creative contribution (prompt engineering + substantial editorial intervention, selection/curation), that human can claim authorship for the protectable elements they originated. Where an AI output is truly autonomous and no human creative contribution is shown, registration and protection are doubtful under current doctrine.
• Moral rights & attribution: Moral rights (right of paternity, right to integrity) attach to authorship. If AI outputs are attributed to humans, questions arise about false attribution or whether moral rights can be enforced against platforms. Indian law will require adaptation here.
• Infringement & training data: Generative models are trained on vast datasets that may include copyrighted works. Whether training constitutes infringing “copying” or is covered by exceptions (e.g., fair dealing under Section 52) is contested. Courts and policymakers are actively considering how to balance innovation, fair use/fair dealing, and creators’ rights.

5. Business & drafting solutions (what lawyers should advise now)

Because the statutory landscape is unsettled, contractual clarity is the immediate — and powerful — remedy:

• Clear assignment/ownership clauses. Contracts should specify whether users, developers, or the AI provider owns outputs. Absent agreement, statutory defaults apply.
• Prompt-engineering and authorship definitions. If a client wishes to claim copyright, contracts should require documented human contributions (time-stamped prompts, editing records) to evidence authorship.
• Indemnities and warranties. Given training-data risks, require indemnities from providers for third-party infringement, and warranties about datasets used.
• Licences from providers. Understand model provider terms (many grant licences that restrict commercial use or require attribution).

6. What law students should watch next

1. Legislative reform in India. Expect policy papers and amendments clarifying whether computer-generated works will receive autonomous protection and, if so, who the “author” will be. The existing Section 2(d) language gives a statutory foothold for the person who causes creation, but lawmakers may refine this.
2. Administrative guidance and registrations. India’s Copyright Office may issue guidance (as the U.S. Copyright Office has) about registration practices for works containing AI components.
3. Key cases abroad and persuasive impact. Decisions like those in the U.S. denying copyright to fully AI-autonomous works are likely to be cited by courts and policymakers in India.
4. Litigation on training data. Cases testing whether model training infringes copyrights — and exploring Section 52 exceptions — will be decisive for how generative models operate in India.

Conclusion — short answer

Under current Indian law copyright most comfortably vests in human authors or in the person who “causes” a computer-generated work to be created (Section 2(d)). Fully autonomous AI outputs without identifiable human creative input remain legally uncertain; international trends and administrative guidance emphasise human authorship. For creators and businesses, careful contracts, documentation of human contributions, and risk allocation are the pragmatic solutions now while litigation and legislation catch up.

Select references (for further reading)

• The Copyright Act, 1957 — Section 2(d) (definition of “author”) and related provisions. (Indian Kanoon)
• U.S. Copyright Office, Copyright and Artificial Intelligence (Part 2: Copyrightability). (U.S. Copyright Office)
• Recent appellate decisions and reporting on Thaler v. Perlmutter (AI authorship litigation). (Institute of Art and Law)
• Bombay High Court orders on removal of AI deepfakes (illustrative of judicial response to harms). (The Times of India)