Authorship of creative works has been a highly debated topic for over 200 years. Till now, assigning authorship to a human was not that difficult as most of the new technologies assisting in creation of copyrightable works, like cameras and computers, were mere tools and humans were the actual brains behind the creation. However, with the recent rapid growth of AI systems and new learning methods, more and more works are being created by machines without any human interference. Copyright offices around the world are unprepared and are not able to acknowledge the importance of these new sources of creativity.

Denial of copyright has resulted in the AI created works to be released into the public domain. There is an urgent need to reinterpret the laws or create new ones to accommodate works created autonomously by nonhumans, as giving exclusive rights over the works would act as a major incentive for the AI developers. This article proposes to reinterpret the terms "employer" and "employee" for the purpose of work made for hire doctrine, rather than redefining the term "authorship" to include non-humans.

Most countries, including USA, UK, and the EU nations, have refused to grant copyright to anyone when there is no active human contribution to the creation of the work. In Naruto v. Slater1, one of the most recent cases to talk about the issues related to human and nonhuman authorship, the Ninth Circuit had to answer the question of animal authorship in photos. In this case, a monkey named Naruto had taken a picture of itself by using a photographer's camera. The court refused to attribute the ownership to the monkey and focused on the wording of the Copyright Act which speaks of a 'person' to be involved in the creative process, and for a work to be copyrightable, a human must be involved in the creative process.

What is the right question? The author is of the opinion that determining copyrightability of AI created works by questioning whether a computer can be called as the author of a work is the wrong place to start with. We should focus more on writings than authors and it is better to ask whether a computer can create a copyrightable work rather than asking if it can be an author. There have been many attempts with different approaches to formulate a method to give protection to works created autonomously by AI systems. Most of them either propose a new sui generis right or amending the copyright laws. Both of these methods require a substantial amount of change to the present legal regime, which is not desirable.


As the developers and the companies owning AI systems have no direct role to play in the creation of works by the AI, it is impossible under the current copyright regime to assign them the copyright. The "Work Made for Hire" (WMFH or WFH) doctrine presents a feasible option to achieve this. According to this doctrine, in case a work is made for hire, an employer must be considered as the author even if an employee had created the work.2 These guidelines can be extrapolated to the AI industry to issue authorship to the programmers or owners. The WFMH doctrine is an exception to the general rule which states that ownership of copyright rests with the person who actually created the work, and as a result it serves as a perfect template to regulate works created autonomously by AI machines. The Indian Copyright Act, 1957, names the employer as author3, as it acts as an incentive and gives them control over the commercial aspect of the creation. Apart from awarding the employer for the investment made by him, a major reason to credit him as the author, instead of the employees, is to establish accountability and responsibility over the actions of the creators. This reasoning would conceptually fit well with AI created works.

Section 17 statutorily recognizes the author of the work to be the first owner of the copyright. However, this provision is subject to certain exceptions. Section 17(a) and (b) talk about situations where the employer is the first owner of the copyright in works created by authors under his employment. Section 17(c) includes all the other types of works created by an author and which are not mentioned in either clause (a) or (b). Where a man employs another to do work for him under his control, so that he can direct the time when the work shall be done and the means to be adopted to bring about the end, and the method in which the work shall be arrived on, then the contract is contract of service. If, on the other hand, a man employs another to do certain work but leaves it to that other person to decide how that work shall be done and what step shall be taken to produce that desired effect, then it is a contract for service. In Beloff v. Pressdram4, it was held that the true test is whether the employee is part of business and his work is integral part of the business, or whether his work is not integrated into the business but is only accessory to it or the work is done by him in business on his own account. In the former case it is a contract of service and in the latter a contract for service.5 When an AI machine creates a work autonomously, it could be treated as work created under contract for service.

There are many questions left unanswered when it comes to implementing the WMFH model on AI created works. Are the works copyrightable to begin with? And if they are not copyrightable, can the employer possess the copyright through the WMFH doctrine? What happens when an AI machine goes beyond the purview of its "employment"? For analysing the Work Made for Hire doctrine, an AI created work should be treated differently than a work created by an employee for a software company. There are no human authors behind an AI created work, whereas in a traditional employeremployee relationship, the employees create the works as per their prior agreement with the employer. The employees create the work with the involvement of the employer.6 The reasoning provided for giving  copyright to employers is to justify the large costs entailed in nurturing talent and gradually creating a copyrightable work.

Implementing this doctrine under the current copyright regime is impossible without a new legislation or amendments to existing laws. Works created autonomously by an AI machine do not come under WMFH as the relationship between the programmer and the AI system is not exactly an employer-employee relationship in the agency sense. According to section 2(d)(vi), an author is someone who caused the computer-generated works to be created. If the courts decide to interpret it liberally or there is an amendment to the Act expanding the definition of author, then it would be possible to grant ownership of copyright in the works created autonomously by the AI machines to those who were responsible for the creation of the AI systems.


Although the WMFH model has its own potential disadvantages, most of these drawbacks can be fixed with a little help from the legislature and the judiciary. Among all the proposed approaches, the WMFH model provides policymakers the least amount of headache, as there is no need to completely overhaul the copyright laws or create a new type of rights for AI created works. This model only requires small amendments to current legislations and a more accommodating interpretation by the courts. It also changes our understanding of computers, as the current legal systems around the world treat computers as a tool and try to find the human behind the machine.

Although there is no intention of labelling them as persons, modern AI systems can act creatively and independently, and thus this model imposes the same rules that regulate works produced by employees. Another problem being solved by this model is that of accountability gap. As the AI machines are seen as employees or contractors, the owner, programmer, or the user is held accountable for any action taken by the AI. This way, the legal system would be able to have a tight control over creative AI systems, and the users of AI systems will try their best to avoid damages, counterfeiting and infringing upon 3rd parties' rights. By reinterpreting the employer-employee relationship for work for hire doctrine, the potential attribution of copyright to non-humans or machines can be avoided as legal (companies) or natural (programmers) persons are given the copyright for works created by an AI. This would help prevent works from being released into the public domain, which is the current position in many countries, including US, as per Naruto v Slater. It ensures that all AI created works have a human author, thus negating the need for a debate over human versus non-human authorship.

Many AI users choose not to file for a copyright, as the general assumption is that their application would be rejected. This results in the failure to reveal participation of AI machines in the creative process. However, in many cases they intentionally don't disclose any contribution made by the AI system, and not properly accrediting the rightful author is a fair ground for annulling a copyright claim.7 By transferring the copyright to a human employer, policymakers can ensure that works created autonomously by AI machines are documented properly and registered with full disclosure of any contribution made by the AI.


1 Naruto v Slater, No 16-15469 (9th Cir. 2018).

2 US Copyright Office, Circular 9: Works Made for Hire (Sep. 2012), <> accessed 13 February 2019.

3 Section 17, Indian Copyright Act, 1957.

4 Beloff v. Pressdram Ltd., [1973] 1 All E.R. 241 (Ch. D.)

5 accessed on 13 February 2019

6 E Jordan and others, 'Employment Relations Research Series 123 Employment Regulation Part A: Employer Perceptions and the Impact of Employment Regulation Executive Summary' < a s s e t s / b i s c o r e / e m p l o y m e n t - m a t t e r s / docs/f/11-1308-flexible-effective-fair-labour-> accessed 13 February 2019.

7 17 USC § 411(a).

The content of this article is intended to provide a general guide to the subject matter. Specialist advice should be sought about your specific circumstances.