Can a machine be an author? India’s Copyright Office has answered: not yet, and not without Parliament’s intervention. As artificial intelligence increasingly generates commercially valuable images, music, text and audiovisual content that can be indistinguishable from human-created works, the question is no longer theoretical: can copyright law recognise a machine as an author, or must authorship remain with a legally recognised person connected to the creative process?

India’s DABUS decision confronts this question directly. The matter concerned A Recent Entrance to Paradise, an artwork generated using DABUS, an AI system created by Dr Stephen L. Thaler, who was named as the owner while DABUS was named as the author. The Copyright Office rejected the application, not for lack of originality (the Registrar found that the work met the threshold under Section 13 of the Copyright Act, 1957), but because DABUS could not be entered as the author under the existing statutory framework.

The decision turns on the distinction between generating a work and holding legal authorship. The Copyright Office did not exclude AI-generated works from protection; instead, it held that authorship must be attributed to a legally recognised person, and that an autonomous AI system cannot occupy that position.

This bears directly on Section 2(d)(vi) of the Copyright Act, which defines the author of a computer-generated work as “the person who causes the work to be created”. The Registrar rejected the applicant’s proposed distinction between an “upstream cause” and an “immediate cause”, holding that the provision looks to the legally recognised person who causes the work to be created, not merely the entity producing the final output. The use of an autonomous algorithm does not, by itself, displace human authorship. On the facts, the Registrar found that Dr Thaler himself was that person, meaning the applicant’s own case identified the statutory author, just not as claimed.

This creates a practical authorship problem for AI-generated works generally: identifying which person, whether the user, programmer, commissioner or another party exercising sufficient contribution and control, meets the statutory test. The one answer the DABUS decision forecloses is that it can be the AI system itself.

The case also separates authorship from ownership. The Registrar accepted that the two are conceptually distinct and may vest in different persons, but that does not permit naming an incapable entity as author while asserting ownership in another person without establishing a valid chain of title. Under Sections 17 to 19, the author is ordinarily the first owner, and any departure requires a statutory exception or valid transfer, a chain that the applicant’s formulation could not satisfy. The Registrar noted that Dr Thaler could still have been recognised as owner had he been correctly identified as author; the defect lay in the particulars submitted, not in an absolute bar.

This distinction between originality and authorship is likely to matter more in future cases involving human-AI collaboration than in cases involving claims of outright machine authorship. A designer who prompts an AI image generator, selects from the outputs and edits the result presents a far stronger case for human authorship than a system operating autonomously with minimal intervention. The degree of human creative input, control and selection is likely to become the decisive factor.

In practice, parties should document the human contribution involved in AI-assisted creation, including instructions, system configuration, selection decisions, iterations and edits, and should address ownership, warranties, disclosure obligations and risk allocation contractually among developers, users, employers, clients and platforms.

Moral rights reinforce why recognising machine authorship would require legislative reform. These rights are inherently personal, protecting an author’s honour and reputation, concepts that an AI system has no capacity to possess.

The Indian decision joins a consistent global trend. In the United States, the same work, A Recent Entrance to Paradise, was refused registration, and the U.S. Court of Appeals for the D.C. Circuit affirmed in March 2025 that the Copyright Act requires eligible works to be authored by a human being. DABUS patent applications met similar rejection across jurisdictions: the United States (affirmed by the Federal Circuit in August 2022), the United Kingdom (Supreme Court, December 2023), the European Patent Office Board of Appeal (2022), and, after an initial acceptance later overturned on appeal, Australia, alongside Taiwan, Switzerland, Germany, Japan and China. This consistency reflects a shared premise that intellectual property frameworks presuppose legally recognised persons, not machines, as creators or inventors.

The Registrar also treated any expansion of authorship to machines as a matter for Parliament rather than administrative interpretation, given the consequences such a change would carry for ownership, licensing, enforcement and moral rights.

India’s DABUS decision is a call for legislative clarity, not a final word. Machines may generate expression that appears original and valuable, but the law continues to look to the legally recognised person who caused the work to be created. Whether a new category of computational authorship will one day emerge remains a question for Parliament, not the Copyright Office.