Author : Giselle Mascarenhas from Amity University Mumbai
I. INTRODUCTION
The rapid advancement of artificial intelligence (AI) has increasingly prompted authors of creative works to be drawn towards using this technology at every step of the creative process. Whether it is to refine their creative expression or overcome artistic blocks, authors are tempted to turn to AI in order to begin, refine and complete their creations leading to a loss of human creativity in the bargain of innovative technology. This has increasingly led to the phenomenon of Human-AI collaborations especially concerning copyright laws. Human-AI collaboration refers to the strategic partnership between human intelligence and AI-technology aimed at harnessing the unique strengths of both entities to achieve superior outcomes. People believe that using AI refines and heightens their nascent talent and creativity. These developments in the creative industry prompt us to revisit the basic principles of copyright laws applicable in this scenario for the copyrightability of these works and the authorship rights eligible.
Firstly, intellectual property laws were introduced to protect and foster human creativity and innovation. Therefore, to protect any kind of work whether through patents, copyright, or designs, the element of human creativity is most essential to grant such a right. Therefore, in most countries only natural persons qualify for copyright protection; completely alienating AI from being an author or co-author as the principle of legal personality being applied to AI is also quite complex and not applicable in the current scenario. Another essential for copyright protection is originality in the expression of ideas, combining the idea-expression dichotomy wherein only the expression is copyrightable and not the idea itself. The uniqueness of the expression is what is protected thus questioning the originality of an expression generated through AI. Afterall, it builds upon existing copyrighted material to create a new expression which may or may not turn out to be completely original, further being difficult to ascertain. Lastly, joint authorship is available only when both authors have substantially contributed to the work however this idea is contradicted by the fact that only natural persons are eligible to qualify as authors.
What about the human mind that has also contributed towards the creation of that particular work? What is the minimum amount of human input that is required in order to classify as copyrightable work as a whole? Despite foreign guidelines regarding copyrightability of works based on human involvement, the distinction between using AI as a tool versus a substitute for human creativity remains unclear. This study therefore, aims to assess the complexities associated with Human-AI collaborative works and the eligibility and implications of such creations under copyright law.
II. THE NOTION OF AUTHORSHIP
Joint Authorship
The copyright regime is mainly designed to protect human authorship and promote creativity that stems from human intelligence. However, neither the Berne Convention nor subsequent international treaties define the term ‘author’, allowing contracting states to define the exact term in their respective legislations. However, an inference from concepts of moral rights and protection based on an author’s nationality suggests the author being a natural or juristic human.
The right of joint authorship is only awarded when two or more people contribute to a work such that these contributions are virtually indistinguishable from one another. Further, each co-author has an independent copyright in the portion of the work attributed to them, and copyright exists equally for all contributing authors. Therefore, if one author wishes to assign the right of publication, the authorization of all co-authors is required. If joint authorship were to be provided to Human-AI generated works, how would this work? Afterall, AI is nothing but technology that uses algorithms without a human mind working behind to provide answers.
Aalmuhammed v. Lee was a notable case where the court held that joint authorship requires someone to be part of the creative team that masterminds the work or supervises its creation and not merely someone who edits or contributes a few lines. Therefore theoretically, an author that uses AI to refine their works in terms of language and delivery is entitled to be the sole author of the work.
Juristic person
Most jurisdictions including US, UK, EU, Japan have rejected AI as having legal personhood. AI is a computer program which acts based on past training, statistical models, and algorithms. As such it lacks the moral compass, contextual awareness, conflict resolution, and critical thinking capabilities that are necessary for an entity to classify as a juristic person. Legal personhood is not a monolith, and granting it to AI would ignore the social, cultural, economic, political, and legal disparities faced by actual human beings, who still struggle to achieve comprehensive civil rights.
Current Scenarios Worldwide
Stephen Thaler v. Shira Perlmutter involved a US court decision affirming that purely AI-generated works without human involvement are not eligible for copyright protection. The court emphasized human authorship being a fundamental requirement for copyright eligibility. Conversely in UK, the case Paintings Generated by AI stated that copyright in the paintings belonged to the artists who created the AI-algorithm, as they were the ones who exercised control over the creative process.
International guidelines also require a certain level of human input and creativity to qualify as copyrightable works. The US Copyright Office emphasizes that if AI is merely used as a tool to enhance human creativity, the resulting work can be copyrightable however, the human contributions must be substantial and creative. Korea and Japan also emphasize the need for human creative input for copyright protection. The Japan Copyright Office states that the determination of whether AI has been used as a “tool” depends on two factors: whether the person had a “creative intention” and whether the person has made a “creative contribution”. Further, the US Copyright Office advises creators advised to document their creative process, including human-created elements and modifications of AI-generated outputs, to support copyright claims.
III. FAIR USE, CONSENT, AND EQUITABLE REMUNERATION
Generative-AI is able to generate new and different content in response to a text prompt entered by the user each time by relying on large language models (LLMs), which are trained by inputting enormous amounts of data gathered from the internet. The more data available, the better the responses generated. The immediate question is whether these training models infringe copyright of the works available on the Internet and as an extension, whether the works created by AI are also an infringement of copyright.
Baker v. Selden laid emphasis on the copyrightable elements of a work and the idea-expression dichotomy. In the debate of whether the AI-training models constitute fair use, courts have overlooked the simple explanation that some acts of reproduction are not copyright infringement because they are non-expressive reproductions of material forms only, similar to Baker’s reproduction of accounting forms.
Furthering this debate, works generated by AI do not fulfil elements of copyright infringement due to the broad concept of idea-expression dichotomy. However, this stands only when the AI-Generated work is a combination of the AI’s large training model and not just a copy of a particular artist’s work. In such cases the work might be considered ‘original’ in the sense that AI has taken inspiration from existing works on the Internet i.e. the idea; and expressed it originally. In Andersen v. Stability AI Ltd., artists Sarah Andersen and others challenged the defendants’ creation and/or use of Stable Diffusion, an AI-software product. Diffusion used plaintiffs’ artistic works as “training images” resulting in Stable Diffusion producing output images “in the style” of those images. The court dismissed the claim of derivative works but held the defendants liable for direct infringement.
The reality is that the use of AI to generate works and their subsequent infringement of copyright is largely circumstantial and this case proved that even precedents are not helpful largely due to the completely differing facts. Thus, if a work generated is substantially similar to an existing work, or the artistic expression and style of a particular artist then it will constitute as copyright infringement. Conversely, if the work is proved to be merely an inspiration from existing works then it will not constitute copyright infringement. However, the legal system is inevitably being asked to clarify the bounds of “derivative works” under IP laws.
What about obtaining a license from the copyright holder to create an AI-generated work which may be substantially similar to the original work? Here AI-generated derivative works or adaptations, will be permissible within the terms of the licensing agreement, provided they display substantial human input. The creator’s consent is key and an equitable remuneration in the form of royalties must be worked out for the creator to reap the benefits of his original creation. Additionally, an AI-generated work with substantial human contributions must be awarded full copyright protection including all moral and economic rights.
IV. LEGAL FRAMEWORKS AND INTERNATIONAL PERSPECTIVES
Internationally, the consensus is that a significant human contribution is required in order to obtain copyright protection for a work. To attract copyright protection, a work must have a substantial amount of human contribution in terms of creativity, originality and other inputs. The term substantial is a measure of the creative input given by the author for the AI to generate a work, which is undoubtedly a matter specific to the facts of each case.
The sufficiency of human contribution in a work may be determined by a five-part test centred on the process of creating the work; primarily targeting jurisdictions with a subjective standard of originality. The steps include assessing the technical aspect of AI, the human decisions made i.e. quantitative and qualitative, the ‘creative distance’ between decisions, expression and finally the originality requirement. Essentially, the AI-user must have given a detailed prompt which showcased their creative autonomy in the making of the piece. The more free and creative choices a user makes during both or one of these phases, the greater chance of obtaining copyright, as the substantiality of human input is prevalent making the work ‘AI-assisted’ as opposed to it being ‘AI-generated’ where the human input is minimal and the author is not the one exercising their intellectuality as a primary source. Thus, the sweat and brow doctrine is explicitly overruled. Internationally, the consensus is that a significant human contribution is required in order to obtain copyright protection for a work.
United States Copyright Office (USCO)
USCO through its report shed clarity on the copyrightability of AI-assisted works. The necessary authorship standard of human creative expression and originality showcased in works must be assessed on a case-by-case basis. Through an example of two prompts, USCO pointed towards the copyrightability of original elements of the sketch (here, the prompt) as evidence of authorship, and any output depicting identifiable elements of this sketch is considered as a derivative work of the sketch’s copyright. Thus, the specific and detailed nature of the prompt is what attracts copyright protection even if the end result, whether an art or a book is AI-generated. The Zarya of the Dawn graphic novel was granted limited copyright protection by USCO wherein the author utilized generative-AI to produce the graphic images for the novel alongside her novel human-generated storyline. The Copyright Office approved protection for the text and arrangement of images, but denied protection for the individual images within the graphic novel as too much “distance” between the user’s input and the output was identified.
EU Copyright Directives
EU does not have any specific guidelines regulating copyrightability of AI-assisted works. Even the AI Act merely speaks about copyright infringement through AI-training models however; by analysing case laws the EU laws do not appear to be an impediment to human authorship and copyright protection in AI-assisted outputs. In 2019, the CJEU clarified the concept of originality stating that, for a subject matter to be considered original, it is both necessary and sufficient that it reflects the personality of its author, representing their free and creative choices. Similarly, the CJEU stated that the subject matter must be the result of the author’s own intellectual creation.
Therefore, works can be considered eligible for copyright protection as long as they bear the imprint of human creativity. An author needs to be a natural person, exercising their creative freedom by monitoring and executing the preparation, execution and finalization of the work, expressing their creative act with a general authorial intent.
Japan Copyright Law
Japanese law defines ‘work’ as a creatively produced expression of thoughts or sentiments, thus limiting copyrightability to works which express an author’s substantial emotions and thoughts in the creative process. In its report, the Japan Copyright Office illustrates that ‘creative intention’ and ‘creative contribution’ are factors used to determine whether AI was merely used as a ‘tool’ in the creation of works. Further, the amount of instructions, number of generation attempts and selection from multiple outputs are some factors used to determine the ‘creative contribution’ of the human entity.
As opposed to other countries, Japan takes a more progressive approach, permitting limited copyright protection if human involvement can be identified, demonstrating a relatively broader recognition of AI-generated content. Although a 100% AI-generated work cannot attract copyright protection, the existence of human creativity is sufficient.
Indian Copyright Law
The Copyright Act includes the phrase ‘the person who causes the work to be created’ with respect to computer generated works in its definition of ‘author’. Although this might imply that the AI-prompt-user gains copyright, it is pertinent to note that a single-line prompt is the equivalent of the ‘sweat and brow’ doctrine which was discarded by the Supreme Court in 2008. Thus, a likely conclusion is that an AI-assisted work with sufficient input and creativity exercised by the person who causes its creation is eligible for copyright protection.
The “Significant Input” test establishes two essential requirements to determine the copyrightability of an AI-assisted work. Firstly, the objective criterion of whether there was human involvement at all in the creation process. Secondly, the extent of human skill, judgment, and labour invested in its creation must be significant enough that the product would be fundamentally different or non-existent without it.
V. SUGGESTIONS AND LEGAL REFORMS
AI is inherently trans-border necessitating the need for harmonized frameworks to address AI-generated works globally. However, countries have already been setting guidelines for copyrightability of AI-assisted works, while India is lacking in this respect. In the absence of such regulations, it is inevitable that many works will be lost to the public domain due to the lack of clarity regarding such AI-assisted works.
First and foremost, it is essential to amend certain definitions and add new ones.
- Defining AI-assisted works as ‘works created by AI and primarily involve human creativity and freedom of thought wherein the work would be fundamentally different or non-existent without such substantial human intervention.’
- Defining AI-generated works as ‘works that have primarily and autonomously been created by AI without any substantial human intervention.’
- Defining substantial human intervention as ‘substantial creative decisions, editing, modifications or detailed prompts made by a human to the AI-generated content and includes:
- The quality and detail included in the AI-prompt;
- The number of attempts made to refine the AI-generated content to the person’s creative vision;
- The use of AI to merely refine or correct a largely human made creative work;
- The selection and arrangement of data from the AI-generated content to create a work.
- Amending the definition of author to add the phrase ‘and includes AI-assisted works.’
Further, the authorities need to clarify the following points regarding AI-assisted and AI-generated works.
- AI-Generated works are not protected under Copyright Law as there is no modicum of creativity and originality exhibited in the work by a human. Moreover, AI is not a legal personality thus it cannot have authorship over such a work generated by it.
- The burden of proof rests on the author who’s work is AI-assisted to prove that it fulfils the criteria of substantial human input. It is encouraged to record the AI-Generated outputs and their respective prompts, divergence of the final product from the initial AI-output and similarity of the final product to author’s existing creative style.
- Mandate transparency concerning use of AI-tools in the creative process, requiring disclosure of AI-involvement in creation of works in copyright applications, which will then be duly verified by the authorities before registration.
Implementing such reforms in the legal framework will allow persons with artistic qualities to display their visions in the form of AI-assisted works while also being protected by Copyright law. Moreover, it will allow the country to maximize the capabilities of AI to drive innovation and technology while refraining from unseasonable restrictions on the works generated by AI.
VI. CONCLUSION
The complex interplay between human creativity and AI within the realm of copyright law, reveal key insights into the challenges and opportunities presented by AI-assisted and AI-generated works. The prevailing legal principle remains that copyright protection fundamentally rests on human authorship and creativity. While AI undeniably offers powerful tools for creative expression, the works produced in collaboration with or entirely by AI reside in a ‘grey area’ demanding careful consideration of existing legalities and need for reforms. The critical element is the substantiality of human creativity exercised in creation of AI-content which makes it AI-assisted. To effectively navigate this rapidly evolving copyright landscape, policymakers should consider providing legal clarity on copyrightability of AI-assisted works and transparency standards on copyright applications. By embracing technological innovation while preserving the fundamental principles of authorship, creativity, and equitable compensation, policymakers can ensure that the copyright system continues to serve as a catalyst for artistic expression and cultural enrichment in the age of AI. Emphasis must also be placed on international cooperation in this space in order to achieve a global, sustainable framework for AI that provides a reasonable level of protection and clarity of standards for the public. Lastly, the emphasis should be on empowering humans to use AI as a tool for artistic creation while upholding the principles of originality, authorship, and equitable compensation that form the foundation of copyright law.