Who Owns an AI Output in India? Developer, Business or User

Indian law answers the ownership question in two steps. It first asks who the author is, and then makes that author the first owner unless an exception or a written assignment says otherwise.

Indian law answers the ownership question in two steps. It first asks who the author is, and then makes that author the first owner unless an exception or a written assignment says otherwise. Generative AI strains the first step, because the people who build the tool, deploy it and use it are usually different. Here is how each of them fares, on the law as it stands after the DABUS order.

Two provisions, two steps

Section 2(d)(vi) of the Copyright Act provides that, for a computer-generated literary, dramatic, musical or artistic work, the author is “the person who causes the work to be created.”¹ Section 17 then makes “the author of a work” its first owner, subject to listed exceptions.² So ownership of an AI output in India turns first on authorship. Contracts come in only afterwards, to move rights from whoever holds them.

On 31 August 2026 the Registrar of Copyrights gave the first detailed reading of section 2(d)(vi), in an application by Dr Stephen Thaler concerning an image generated by his AI system, DABUS.³ Part 1 of this series covered what the order decided. Three points from it govern the ownership question.

• The inquiry is “an inquiry into legally attributable causation,” not into which machine produced the final output.⁴
• It is specific to each work: “There must be a legally sufficient connection between the acts of the identified person and the origination of a particular work.”⁵
• The author is the person who “exercised such creative control and made such proximate and material arrangements that the work can fairly be regarded as having been caused by that person.”⁶

The order did not have to choose between different people. Dr Thaler built the system, supplied the inputs and started the run, and no text prompt was used.⁷ The order says it is not ruling on every form of AI-assisted creation, because “the degree and character of human involvement may vary materially from one system and one work to another.”⁸ What follows applies the order’s test to the common commercial arrangements. It is our reading; no Indian court or Registrar has yet decided these cases.

1. The developer of the AI tool

A company that builds a general-purpose model is unlikely to be the author of what its users generate. The order is direct on this: “A person does not become the author of every output merely because that person owns, designed or developed the computer system.”⁹ Owning the machine does not help either. The Registrar held that ownership of a computer, software or other instrument does not, without more, carry copyright in its outputs.¹⁰ A developer would need work-specific acts, such as supplying the inputs or directing a particular output, to come within the test.

Some developers are not seeking these rights anyway. In the UK consultation, OpenAI said it does “not claim copyright over generated outputs.”¹¹

2. The user who writes the prompt

This is the open question. The user is the person closest to the particular output, which fits the order’s emphasis on the specific work. But the order also asks for “creative control” and “proximate and material arrangements,” and it has not said whether typing a prompt meets that standard.

Other systems point in different directions. The UK government reads its own provision as making the prompter “usually” the author of a general-purpose AI’s output.¹² The U.S. Copyright Office takes the opposite view of authorship for US purposes: “prompts alone do not provide sufficient human control to make users of an AI system the authors of the output.”¹³ India’s test sits between the two. A user who selects, arranges or edits outputs is on firmer ground than one who accepts the first result of a one-line prompt.

3. The business whose employee uses the tool

If the employee is the person who caused the work to be created, section 17(c) ordinarily makes the employer the first owner of a work made in the course of employment under a contract of service, in the absence of an agreement to the contrary.¹⁴ That route works without any new law.

A harder question is whether the business itself can be the “person” who causes the work. The Act’s references to a person include juristic persons such as companies where the statute permits,¹⁵ and the amicus in DABUS suggested that authorship vests in the legally recognised person who “conceived, configured, deployed or otherwise caused the generative process to occur.”¹⁶ The Registrar did not decide the point. Until a court does, businesses should rely on section 17(c) and their employment contracts rather than on corporate authorship.

4. The agency working for a client

Section 17(b) makes a commissioner the first owner of “a photograph taken, or a painting or portrait drawn, or an engraving or a cinematograph film made, for valuable consideration” at the commissioner’s instance.¹⁷ It is tempting to apply this to AI images made to a client’s brief. The words fit AI images poorly, though, and the amicus in DABUS stressed that the provision applies only to “specified categories of commissioned works.”¹⁸ The safer route is an express assignment.

Indian assignments have formal requirements that foreign templates often miss. An assignment is not valid unless it is in writing and signed by the assignor or an authorised agent.¹⁹ It must identify the work and specify the rights assigned, their duration and territorial extent, and the royalty or other consideration.²⁰ Where the duration is not stated, it is deemed to be five years. Where the territory is not specified, it is presumed to be India. Rights not exercised within one year are deemed to lapse unless the assignment says otherwise.²¹

Practical points for foreign counsel

• Paper the Indian chain of title. For AI-assisted work created in or for India, use signed written assignments that meet section 19, with stated duration and territory, and express wording on the one-year lapse rule.
• Record who did what for each work. Authorship under the order follows work-specific acts. Keep a record of who supplied the inputs, configured the process, and selected or edited the output.
• Do not rely on owning the tool or on platform terms. Owning the system does not confer copyright.²² A tool’s terms can allocate whatever rights exist between the parties, but in our view they cannot make someone an author under section 2(d)(vi).
• Treat prompt-only outputs as uncertain. Neither authorship nor originality has been settled for them in India.

Footnotes

1. Copyright Act, No. 14 of 1957, § 2(d)(vi) (India) (inserted by Copyright (Amendment) Act, No. 38 of 1994, § 2, w.e.f. May 10, 1995). ↩

2. Id. § 17. ↩

3. In re Diary No. 9356/2022-CO/A, ¶¶ 1, 189–190 (Registrar of Copyrights, India Aug. 31, 2026) [hereinafter DABUS Order]. ↩

4. DABUS Order ¶ 91. ↩

5. Id. ¶ 98. ↩

6. Id. ¶ 127. ↩

7. Id. ¶¶ 16, 107. ↩

8. Id. ¶ 23. ↩

9. Id. ¶ 98. ↩

10. Id. ¶ 165. ↩

11. Dep’t for Sci., Innovation & Tech. & Dep’t for Culture, Media & Sport, Report on Copyright and Artificial Intelligence § I, ¶ 37, at 109 (Mar. 2026) (U.K.) [hereinafter UK Report]. ↩

12. UK Report, supra note 11, § I, ¶ 1, at 104; see Copyright, Designs and Patents Act 1988, c. 48, § 9(3) (U.K.). ↩

13. U.S. Copyright Office, Copyright and Artificial Intelligence, Part 2: Copyrightability 18 (2025). ↩

14. Copyright Act § 17(c). ↩

15. DABUS Order ¶ 125. ↩

16. Id. ¶ 78(a) (submission of the amicus curiae). ↩

17. Copyright Act § 17(b). ↩

18. DABUS Order ¶ 148 (submission of the amicus curiae). ↩

19. Copyright Act § 19(1). ↩

20. Id. § 19(2)–(3). ↩

21. Id. § 19(4)–(6). ↩

22. DABUS Order ¶ 165. ↩