The Ambiguity Of Copyrightability Of AI-Generated Works

To understand the crux of the legal issues, it is imperative to know how the generative AI model works. Contrary to popular misconception, generative AI does not create content ex nihilio. While it does generate outputs, it does not do so with absolute autonomy, rather, its operations are intrinsically reliant upon the content or data upon which it trains.

Generative AI trains by feeding massive content in the form of datasets via a process called machine learning, these datasets comprise diverse content ranging from text and images to audio and video files. After the user inputs text prompts or commands, the AI model analyses the input, processes it using its trained parameters, and then predicts the desired output.

The bulk of data used for training AI models is primarily sourced from the internet, albeit the sources are not limited to that. Since the amount of content is enormous, there is always ambiguity as to whether the data used for training was subject to copyright protection, and if so, whether the developers duly obtained the requisite permissions or licenses from the respective authors for its utilization. Since companies generally do not disclose these specifics, in addition to the lack of a legal mandate requiring them to do so, the possibility of copyright infringement in the development of generative AI is high. To escape from such legal liability, Companies rely on the concept of Fair Dealing. This concept is one of the instruments that incentivize creators while also fostering a thriving public domain of idea and expression. In India, “fair dealing” has not been defined specifically, but it has been incorporated in the Copyright Act, 1957,^[1] which permits limited use of copyrighted works for the purpose of research, criticism and reporting without acquiring permission from the creator. Finally, the question remains, whether a work created by such a method can pass for copyright protection?

Classification Of AI-Generated Works: Original Or Transformative?

An analogy can be drawn between AI and an artist who, after finding inspiration from works of multiple artists, creates fresh art pieces driven by newfound knowledge. In a similar vein, AI-generated creative works do not entail direct replication of pre-existing content but rather result from "learning" and implementing available data. Such works, despite not being completely original, may possess the necessary transformative nature that may make them eligible for copyright protection.

When the matter of copyrightability of these AI-generated works is presented before the appropriate authorities, it becomes crucial to examine and understand the fundamental functioning of the specific AI tool involved and carefully assess the creative work in question. By comprehending the underlying mechanisms of the AI and scrutinizing the degree of transformation of the generated content, a more informed determination can be made regarding its eligibility for copyright protection.

Transformation in creative works varies on a case-to-case basis. In R.G. Anand v. M/S. Delux Films & Ors^[2] the Supreme Court of India held that substantial similarity between two works should be avoided, to the extent that a reasonable spectator, upon viewing both works simultaneously, would not conclude one as a mere copy of the other. In East Book Company & Ors v. D.B. Modak & Anr^[3] the Supreme Court pronounced that for any work to receive copyright protection, it must be demonstrated that it is more than just a copy of the original and must contain the author’s independent work.

Authorship Of AI-Generated Work: Indian Perspective

The question of conferring authorship to generative AI has not been expressly addressed in the Copyright Act, 1957 since the law recognizes the author to be a legal person, and AI has not(yet) been interpreted to fall within the ambit of “person”. Although, the Copyright Act does recognize the author of a computer-generated literary, dramatic, musical or artistic work to be the person who causes the work to be created,^[4] which can be interpreted on a factual basis to grant authorship to the person whose intervention led to the creation of the generative work. Since the AI and copyright conundrum is fairly new in Indian perspective, enough precedents are not available to concretize a uniform legal standpoint.

The most notable case pertaining to this question surfaced when an application was made to grant copyright to AI RAGHAV for generating the artwork titled “Suryast”. The application was initially rejected but later, another application was made where RAGHAV and its creator Ankit Sahini were named as co-authors and received registration on 2 November 2020. A year later, a withdrawal notice was issued by the Copyrights Office, but no final decision has been taken in this case and the artwork remains listed as registered.^[5]

Global Trend: UK Vs US

Further, in dealing with the question of copyrightability and authorship of AI-generated work, it becomes important to consider the different approaches taken at the global level. The English approach has been far more lenient than its American counterpart. The US does not extend copyright protection to computer-generated works without a human author, while discussing the application for copyright registration of a comic book in which the AI tool Midjourney was named as a co-author, the US Copyright Office (“USCO”) discussed the eligibility of such authorship and reiterated that:

“The term “original” in this context consists of two components: independent creation and sufficient creativity. First, the work must have been independently created by author. Second, the work must process sufficient creativity.^[6]”

Further, the Compendium of USCO Practises also states that:

“The copyright law only protects “the fruits of intellectual labour” that “are founded in the creative powers of the mind. Because copyright law is limited to “original intellectual conceptions of the author,” the Office will refuse to register a claim if it determines that a human being did not create the work.^[7]”

Such instances clarify that the US does not confer authorship to AI. Whereas, UK’s Copyright Design and Patents Act (CDPA), 1988 expressly provides authorship of computer-generated work (AI included) to the person who made the necessary arrangement for the creation of the work^[8] and confers copyright protection to it for a span of 50 years^[9].

The Way Forward

AI is in its infancy, but its possible negative repercussions are however causing growing concerns. The growing dependence on AI tools calls for the attention of lawmakers. Considering this, IP protection in the age of AI needs to be formed in ways that are helpful to humanity. Even in cases where the output of an AI system is unique and innovative and hence deserving of copyright protection under current legal frameworks, the crucial issue of authorship and, consequently, ownership will depend on which human provided the necessary creative input or intellectual effort.

Drawing on recent rulings, a "clear and stable legal framework" is a pressing priority, however, as a substitute, a sui generis system for regulating AI can be considered which may cover all aspects of AI, including copyright. This could be a good option given the requirement to provide legal clarity and a reliable legal environment for A.I.

Views are personal.