"Guiding the Machine: Assessing Human Authorship in AI-Generated Works for Work-Made-for-Hire"

Abstract

The rise of Artificial Intelligence (AI) as a content creator challenges the established “work-made-for-hire” doctrine in Indian copyright law. This doctrine, enshrined in Section 17 of the Copyright Act, grants ownership of copyright to employers for works created by employees during their course of employment. However, the concept of authorship, traditionally attributed to humans, becomes ambiguous when AI generates content.

This paper explores the potential impact of AI-generated content on the work-made-for-hire doctrine in India and aims to answer the following question:

1. Does the human user providing prompts and guidance to the AI tool qualify as sufficient authorship to trigger work-made-for-hire?

This research paper explores whether the involvement of a human user in providing prompts and guidance to an AI tool qualifies as sufficient authorship to trigger the “work-made-for-hire” doctrine under copyright law. The study examines the legal definitions and interpretations of authorship and work-made-for-hire, alongside the evolving nature of AI-generated works. Through a comprehensive analysis of case law, including 2023 case of Stephen Thaler v. Shira Perlmutter, Register of Copyrights and Director of the United States Copyright Office, et al and Supreme Court of India’s 2008 judgement in Eastern Book Company v DB Modak, the relevant statutory provisions, and academic commentary, the paper investigates the extent to which human input influences the authorship status of AI-generated content. A clear legal framework will be crucial for fostering innovation in the Indian AI landscape while ensuring proper attribution and ownership of copyrighted works.

Keywords : AI, Authorship, Work for hire doctrine, Copyright law, Employee,

Introduction

AI algorithms can now autonomously generate a myriad of creative works, ranging from articles and music compositions to visual art and even entire novels. This unprecedented capability challenges conventional notions of authorship and raises concerns about the applicability of existing copyright doctrines. For instance Grammarly’s “CoWrite” feature uses AI to brainstorm ideas and write alongside you.

Another example can be Vocaloid, a software developed by Yamaha, that allows users to synthesise singing by inputting lyrics and melody. It uses voice samples and phonemes to create realistic vocal tracks, enabling users to produce songs with virtual singers. In this scenario, where there is no direct human involvement in the creative process, determining copyright ownership becomes increasingly complex, blurring the lines between creator and machine.

In the case of Stephen Thaler v. Shira Perlmutter, Register of Copyrights and Director of the United States Copyright Office, et al, the US district court held, the Copyright Act requires human authorship. As explained by the Court, an “author” is “an originator with the capacity for intellectual, creative, or artistic labour,” which is necessarily a human being.

In another ongoing case between Getty Images and Stability AI, Getty Images, a giant in the stock photo industry, accuses Stability AI of copyright infringement. They allege that Stability AI’s image generation software, Stable Diffusion, infringed on their rights by using millions of images, captions, and data from Getty’s collection to train its AI model. This training process, according to Getty, essentially copied their intellectual property without permission. Furthermore, Getty argues that Stable Diffusion’s ability to generate images containing watermarks similar to theirs constitutes trademark infringement. Stability AI has yet to fully respond to the accusations. The core legal question revolves around how copyright applies to training AI models and whether the AI-generated images themselves are too similar to the originals, infringing on their copyright. A decision in favour of Getty Images could set a legal precedent requiring stricter control over how copyrighted data is used to train AI models. Conversely, a win for Stability AI could open the door for broader use of copyrighted material in AI development. The ultimate outcome of the case will have significant implications for the future of AI art and the boundaries of copyright in the digital age.

Indian copyright law, similar to other jurisdictions, revolves around human authorship. AI can be programmed to generate creative outputs like artwork, music, or text without any human creator in the traditional sense. Thereby When AI creates a work, it’s unclear who the legal author is – the programmer, the person who commissioned the AI, or the AI itself? This ambiguity makes it difficult to determine who owns the copyright under the “work made for hire” framework. Leading to disputes between the commissioner and the programmer regarding copyright ownership of the AI-generated work. As AI systems create artwork independently, it becomes imperative to analyse the implications of this emerging trend on copyright, ownership, and the very definition of creativity.

Copyright protection also requires a certain level of originality. While AI-generated works can be impressive, the question arises – does the AI’s output meet the originality standard set by Indian copyright law? If not, the “work made for hire” concept becomes irrelevant.

And lastly If AI can create copyrightable content, there might be less demand for human creators working under the “work made for hire” model, potentially affecting their livelihood.

The current legal framework is inadequate to adapt to these new challenges. Till new laws are made there is a need for sui generis approach to fill the lacuna in law. As has been long accepted, only a human being can be considered an author. Thus to incorporate AI under the copyright laws the definition and interpretation of the term “employee” needs to be stretched.

There are 4 questions needed to be answered to understand the impact of AI-generated content on the “work-made-for-hire” doctrine, where an employer owns the copyright of works created by an employee? Specifically, whether AI can be interpreted as an “employee” under the work made for hire doctrine, whether the human user who is providing prompts and guidance to the AI tool qualify as sufficient authorship to trigger work-made-for-hire? Can AI, lacking legal personhood, be considered an “author” under Indian law? If not, the work-made-for-hire doctrine might not apply, potentially leaving copyright ownership unclear and finally When humans and AI collaborate on creative projects, who owns the copyright?

This paper specifically deals with the question, whether the human user who is providing prompts and guidance to the AI tool qualify as sufficient authorship to trigger work-made-for-hire?

Research Methodology

For the purpose of this paper, a combination of Secondary and Primary research has been adopted. Various case laws, statutes, Newspaper articles, and blogs were reviewed for this paper.

Review of literature

The Work-Made-For-Hire Doctrine

The work for hire or work made for hire doctrine defined under Section 17 of the Indian Copyrights Act of 1957 under the category of First Owner of the Copyright.According to clause 17(c) of the Indian Copyrights Act of 1957, “in the case of a work made in the course of the author’s employment under a contract of service or apprenticeship, to which clause (a) or clause (b) does not apply, the employer shall, in the absence of any agreement to the contrary, be the first owner of the copyright therein;” this means that if an employee creates a work during their regular course of employment under a service contract, the employer will be deemed the first owner unless there is a contrary agreement.

Similarly section 17(b) applies to commissioned works, it states that “subject to the provisions of clause (a), in the case of a photograph taken, or a painting or portrait drawn, or an engraving or a cinematograph film made, for valuable consideration at the instance of any person, such person shall, in the absence of any agreement to the contrary, be the first owner of the copyright therein;” thus when someone commissions the creation of a specific work, like a photograph, painting, or film, for a fee, in the absence of a written agreement stating otherwise, Section 17(b) grants the copyright ownership of the work to the commissioner, not the creator.

In Khemraj Shrikrishnadass v. M/s Garg & Co. (1975), the Delhi High Court ruled that, unless there is a specific arrangement to the contrary, copyright is frequently transferred to the publisher when one author completes work for another writer in exchange for compensation. It is up to the contracting parties to find out how to break out of such commitments. Due to the lack of contractual duty, freelancers are considered the original proprietors of copyright in Indian law, whereas periodicals, journals, and newspapers are considered the original works created by workers under a service contract.

Many prominent modern artists, such as Andy Warhol, Kehinde Wiley, and Louise Bourgeois, are recognised as the creators of works that they did not actually produce. In 1966, Cavalier Magazine posed a question to Andy Warhol on the uniqueness of each of his pictures. Warhol said, “Why don’t you ask my assistant, Gerry Malanga, some questions? He did several of my paintings.”

One of the wealthiest and most well-known painters alive, Damien Hirst, is the source of another well-known example. His 1400 spot paintings were nearly entirely produced by apprentices in Hirst’s workshop. Hirst was reported as stating, “The best person who ever painted spots for me was Rachel; the spots I painted are terrible,” in an article by The Independent.

These are classic cases of work for hire doctrine, where even though the paintings were made by someone else the credit and ownership rights were given to the employer.

But the issues arise when an employee creates intellectual property after hours while utilising company resources; after hours without utilising employer resources; during working hours while utilising personal resources; or unrelated to his job during company working hours.

In these instances, the courts have relied on the “duty to invent” or the “nature of work” to determine ownership rights in the absence of a contractual agreement. Therefore, whether or not the work is produced using personal resources or outside of working hours, it belongs to the employer if it is produced by an employee in accordance with the terms of his employment contract and related to the nature of his duties. However, even if intellectual property is developed during working hours, it still belongs to the employee if it has nothing to do with the job for which they were engaged.

The US courts have developed certain tests to determine whether an employer-employee relation exists between the parties. These tests have been widely recognised by other countries including India.

1. The Right to Direct and Control Test:

This approach focuses on the hiring party’s authority to control the work’s creation, regardless of whether they exercise that right. If the hiring party has the power to dictate the work’s direction and execution, the creator might be considered an “employee”.

2. The Agency Law Test:

This test applies principles of agency law to assess the relationship between the parties. Courts consider several factors to determine if the hired party functions as an agent, and therefore a statutory employee.

3. The Actual Direct and Control Test:

This test emphasises actual supervision and control exercised by the commissioning party during the work’s creation. If the commissioner actively oversees and directs the work, it strengthens the argument for work made for hire.

4. Salaried Employee Test:

This defines an “employee” based on traditional employment characteristics like regular salary, tax withholding, and benefits.

Thus When a client asserts ownership of your work as a product of a work-for-hire agreement, you, as the creator, relinquish your copyright from the outset. One significant ramification of this doctrine is the loss of control over your work’s usage. The client gains the authority to publish the work at their discretion, including the option to sell rights to third parties without your consent. Without explicit permission from the client, you are restricted from utilising your own work in any capacity.

AI and the Concept of Authorship

The traditional definition of authorship in copyright law revolves around the concept of human creativity and originality. According to this definition, an author is typically understood as the individual or group of individuals who create an original work fixed in a tangible form. This concept emphasises the role of human intellect, creativity, and effort in producing something new. The law recognizes the labour, skill, and talent involved in the creation of original works, whether they are literary, artistic, musical, or other forms of expression.

Under this framework, copyright law traditionally grants exclusive rights to authors to control the reproduction, distribution, and adaptation of their works for a certain period of time. These rights are meant to incentivize creators to continue producing new works by providing them with a form of legal protection and the potential for financial reward.

However, as technology continues to advance, questions arise about whether the traditional definition of authorship adequately addresses new forms of creative expression, such as computer-generated works or collaborative projects involving AI systems. These developments challenge traditional notions of authorship and raise complex legal and ethical questions about ownership, attribution, and the nature of creativity in the digital age.

A large amount of text, photo, audio, and video data—much of it is taken from the internet—is used to train large language models in artificial intelligence. The author’s guild believes that texts are copied in the software which is then reproduced again and again. However as per Matthew Sag, professor of law in artificial intelligence, machine learning and data science at Emory University, rather than copying the data it makes more sense to view the software as a student being trained based on the data.

However the bottom line remains that the AI software is trained by material available on the internet, some of which might be copyrightable. So the question arises for future research whether this would amount to fair use exception or would this be considered an infringement.

The earliest cameras were introduced in the 1880s. Oscar Wilde was later photographed by Napoleon Sarony, whose images were printed without permission by the Burrow-Giles Lithographic Company. Sarony filed a lawsuit alleging infringement of copyright. The corporation responded by saying that people who use cameras only utilise equipment to replicate reality, not to create creative pieces of art. The US Supreme Court, however, disapproved.

According to American photographer Stephen Shore, who is mostly recognised for having pioneered colour photography as an art form, “it means something that three-dimensional space is flattened into two-dimensional space, that time is stopped.” “Those who are aware of this can use a camera as a tool to make their own creations that are distinct from the outside world.” Content produced by AI may likewise be justified using this logic.

“It is an artistic, iterative process,” Mahari stated. “However, it is unclear where the line is drawn… How originality and creative excellence are seen in a society where reasonably high-quality artworks are relatively easy to generate is still being debated.” The law is less clear about who owns AI outputs. Certain writers believe that they hold the rights to the outputs generated by AI systems since their copyrighted works are utilised to train the robots. Others argue that those who use AI as a tool are the writers.

AI and Work-Made-For-Hire

Assigning authorship to an individual has been relatively easy up until now because the majority of modern technology, like computers and cameras, that aid in the production of works protected by copyright, are only tools and the actual minds behind the creation are humans. But as artificial intelligence (AI) technologies and new learning approaches proliferate, more and more works are being created by computers independently of humans. It is difficult to pinpoint the author in these situations.

AI algorithms, such as GANs, deep learning models, Natural language processing (NLP) algorithms, are used to create original artworks, including paintings, sculptures, digital graphics, multimedia installations, articles, stories, poems, novels, musical compositions, visual effects, prototypes, etc.. . Some AI systems can even autonomously create art without human intervention. AI-powered tools can analyse and study various works and mimic their styles or create new works based on themes or prompts, producing outputs that closely resemble human work, fostering collaboration between human and AI-generated ideas, based on user preferences and input.

Imagine the following: every day, hundreds of people visit the Eiffel Tower. Even though they may not have had much input, every person who snapped a picture of the Eiffel Tower is acknowledged as the photographer of their own image. With the exception of pressing the button, the camera does the most of the real work in creating the image of the Eiffel Tower.

Similarly, one may instruct an AI to draw a picture of the Eiffel Tower. In the same way that a photographer has little to no engagement in the process other than pressing a button, the user has little to no input in the actual creation of the piece.

The Naruto v. Slater case often referred to as the monkey selfie case, garnered media attention recently in the US. Here, a monkey named Naruto used Slater’s camera—a wildlife photographer—to snap several photos, including selfies. The monkeys were interested in the equipment, so Slater set up his camera on a tripod and took a few pictures. He claims he then adjusted the camera settings. After that, the monkey continued to snap several photos. The images were subsequently forwarded by Mr. Slater to National Geographic and other media organisations. The campaign organisation People for the Ethical Treatment of Animals (PETA) filed a lawsuit against Mr. Slater in 2015, claiming that the selfie was the product of “a series of purposeful and voluntary actions by Naruto, unaided by Mr. Slater, resulting in original works of authorship not by Mr. Slater, but by Naruto”.

The trial court rejected the complaint in 2016 on the grounds that, even if the images were an “independent, autonomous action,” animals lacked the legal standing to file a copyright infringement lawsuit. PETA filed an appeal with the 9th Circuit Court of Appeals, but an out-of-court settlement was struck in which Mr. Slater committed to provide the monkey sanctuary in Naruto twenty-five percent of his future royalties.

In these cases even though the work was mostly done by technology, the right to authorship of the person initiating it is recognised. Thus this analogy can also be extended to the person who initiates or prompts the AI to do a work for him.

In certain Indian court cases, such as Eastern Book Company v. DB Modak (2008) 1 SCC 1 and the Delhi High Court Division Bench in Dart Industries Inc. v. Techno Plast (2016), the courts have held that not every endeavour, skill exercise, or industry produces a work that is protected by a copyright; rather, only works whose creation requires some amount of intellectual labour and creativity can be afforded copyright protection.

Thus if any work is created by AI, to give authorship to the person initiating it there needs to be some element of creativity for it to be copyrightable. And if the person is an employee and the work is done in relation to his contract of service then the ownership will be with the employer.

Conclusion

The rise of Artificial Intelligence (AI) presents a fascinating challenge to copyright law, particularly regarding ownership and the work-made-for-hire doctrine. While current interpretations suggest human authorship remains essential, the lines blur as AI capabilities become more sophisticated.

For India’s burgeoning AI landscape to flourish, a clear legal framework regarding AI and copyright is crucial. Uncertainty stifles innovation, hindering both the development of AI tools and the exploitation of AI-generated creative works. A framework that balances the rights of creators (human or otherwise) with the need to foster innovation is essential.

Suggestions

To ensure proper attribution and ownership of copyrighted works involving AI, several steps can be taken:

Transparency: User agreements for AI tools should explicitly address copyright ownership and attribution. Did the user provide sufficient creative input to be considered an author, or does ownership solely reside with the AI tool’s developer?

“Collective Authorship”: In some cases, “collective authorship” might be appropriate, recognizing the contributions of both the human user and the AI tool.

Legislative Updates: The Indian Copyright Act might need revisions to address the unique challenges posed by AI-generated works. This could involve defining new categories of authorship or adapting the work-made-for-hire doctrine to the AI era.

By fostering a clear legal framework and encouraging responsible practices, India can ensure its AI sector thrives while protecting the rights of all creators, human or otherwise.

References

(n.d.).US Department of Justice. https://www.justice.gov/osg/brief/easter-seal-society-crippled-children-and-adults-louisiana-inc-petitioner-v-playboy
Allyn, B., ‘New York Times’ considers legal action against OpenAI as copyright tensions swirl. (NPR 2023, August 16).
Dart Industries Inc & Anr. vs Techno Plast & Ors. on 21 July, 2016.
D’Agostino, S. AI raises complicated questions about authorship. Inside Higher Ed. (2023, August 22). https://www.insidehighered.com/news/tech-innovation/2023/08/22/ai-raises-complicated-questions-about-authorship
Getty Images v Stability AI: copyright claims can proceed to trial. Pinsent Masons. (2023, December 11). https://www.pinsentmasons.com/out-law/news/getty-images-v-stability-ai
Johnson, A. A Damien Hirst original… | The Independent. (2008, September 14). https://www.independent.co.uk/arts-entertainment/art/news/a-damien-hirst-original-929872.html
Naruto v. Slater, No. 16-15469 (9th Cir. 2018)
The Copyright Act, 1957. Section 17.No. 14, acts of Parliament, 1957
Shuman, C. Copyright Office Reiterates that Works Created by AI Cannot be Copyrighted. Pearl Cohen. (2022, February 22). https://www.pearlcohen.com/copyright-office-reiterates-that-works-created-by-ai-cannot-be-copyrighted/
Smith, J. Creativity and Technology: Exploring AI Authorship. AJE (2023, September 29) https://www.aje.com/arc/ai-and-authorship/
THALER v. PERLMUTTER (2023)

By Priyal Thakkar

NMIMS’ Kirit P. Mehta School Of Law

“Guiding the Machine: Assessing Human Authorship in AI-Generated Works for Work-Made-for-Hire”

Contents