The increasing use of Generative AI, i.e. the branch of artificial intelligence dealing with systems capable of producing original content (text, images, sound and synthetic data), without direct human intervention, gives rise to several copyright infringement concerns.

This article focuses on the rights of creators whose work has been utilized to train generative AI models and whether content generated by an AI system is eligible for copyright protection.

Generative AI systems are trained to produce outputs by being exposed to considerable pre-existing creative content and artistic works. OpenAI claims that it collects a large amount of data like images, sentences, or sounds, etc., from some domain and then trains an AI model to generate similar data. It is imperative that such training of generative models does not lead to the replication or creation of derivative works from the original data, however, since the process involves copying the source data, there exists uncertainty regarding the extent to which such training constitutes violation of the original content creator’s copyright.

The existence of Generative AI presents a challenge to the implementation of fair use principles under Section 52 of the Indian Copyright Law, that is an exemption to the exclusive rights of a copyright owner, including inter alia the private or personal use of any work for research, criticism, review, reporting and utilising a computer programme for its intended purpose, making temporary back-ups as a safeguard and for non-commercial personal use. Similarly, in the United States, fair dealing is determined based on factors including the purpose of the usage; substantiality of the portion of the original work used; the effect upon the relevant market and nature of the copyrighted work. In the case of Campbell v. Acuff Rose Music[1], the court emphasised that the exception of fair use strengthens when the new work is “transformative”, modifying the original work with new expression or meaning. Further, in Andy Warhol v. Goldsmith[2], the court held that if a secondary work, being used commercially, shares a highly similar purpose as the original work, establishing fair-use becomes challenging unless compelling justifications exist.

Thus, utilization of original works for training, enables AI systems to generate independent outputs, which can be considered as transformative, providing the output does not violate the original content. However, when considering whether the act of commercial copying of works into training data qualifies as a transformative fair use, the same may fail to meet the criteria of fair use. In the case Andersen v. Stability AI et al[3], filed in 2022 in the United States, the judge ruled that the plaintiff artists arguing that Generative AI utilized their original works for training AI systems, must demonstrate that derivative works generated by AI programs in response to user prompts, bear substantial similarity to their original works.

The subsequent apparent concern is whether AI generated content is capable of copyright protection. In the United Kingdom, only original works which are the ‘intellectual creation’ of the author is eligible for copyright protection as held in the case of Infopaq International A/S v Danske Dagblades Forening[4]. Further, in a recent case of Thaler v Comptroller-General of Patents, Designs and Trademarks[5], the Appellant filed grants for two patents inventions where the inventor was an AI powered machine, however, the court held that an inventor must be a natural person and denied the grants of patents to the AI system. The United States also does not extend copyright protection to works that are solely machine generated. However, in China, in the case of Li v Liu[6], the court held that AI-generated pictures met the criteria for copyright protection under Chinese law, though, the authorship of the AI generated image was attributed the plaintiff due to his intellectual input to the process.

In India, Section 2(d) of the Copyright Act, 1957 defines an author as any person who causes the work to be created. In 2020, the Indian Copyright Office rejected an application for an AI generated artwork where the sole author was an AI tool named RAGHAV, yet granted registration to another artwork applied in the name of the same AI tool along with a natural person. However, the Copyright office later issued a Withdrawal Notice to the Applicant over the uncertain legal status of the AI listed as the co-author. Thus, the extension of copyright protection to content generated by AI remains uncertain in India.

The concerns of copyright protection and artificial intelligence systems were also addressed in the 161st Parliamentary Standing Committee Report[7], which found that the Copyright Act, 1957 is not sufficiently equipped to facilitate authorship and ownership of intellectual property by AI and recommended that a separate category of rights for AI and AI-related works should be created for their protection. The report further suggested for the Copyright Act to be revisited to incorporate AI generated works in its ambit. The Ministry of Electronics and Information Technology is also taking steps to address issues arising with the advent of Generative AI and has advised that platforms in India that are testing or training AI tools, will need to seek permission from the government before launching the product or making the same available to users in India. The Ministry has also proposed a Digital India Act, to supersede the Information Technology Act, 2000, with the objective of regulating Artificial intelligence systems and providing protection to the rights of news publishers and content creators in their original works.

Thus, the potential for AI systems to autonomously create content that closely resembles existing works challenges traditional notions of Copyright Law, based on human authorship and creativity. It is essential for policymakers and technology developers to establish clear guidelines and regulations to address gaps in the existing legal framework for copyright protection, as well as develop mechanisms to attribute authorship and ownership of such Generative AI content, that has several use cases, in the spheres of healthcare, agriculture, governance, amongst others.

Author: Christine Chiramel

Footnotes

[1] Campbell v. Acuff-Rose Music, Inc., 510 U.S. 569 (1994); https://supreme.justia.com/cases/federal/us/510/569/

[2] Andy Warhol Found. for the Visual Arts, Inc. v. Goldsmith 143 S. Ct. 1258 (2023); https://www.supremecourt.gov/opinions/22pdf/21-869_87ad.pdf

[3] Andersen et al v. Stability AI Ltd. et al, No. 3:2023cv00201; https://law.justia.com/cases/federal/district-courts/california/candce/3:2023cv00201/407208/166/

[4]  https://eur-lex.europa.eu/legal-content/EN/TXT/?uri=CELEX%3A62008CJ0005

[5] https://www.supremecourt.uk/cases/uksc-2021-0201.html

[6] https://www.dwt.com/blogs/artificial-intelligence-law-advisor/2024/01/chinese-internet-court-rules-on-ai-authorship#:~:text=Liu%20had%20violated%20Mr.,Li.

[7] https://www.tbalaw.in/_files/ugd/f4c767_94ced94c2ebe42df952d8349f43d28f0.pdf

More from Rahul Chaudhry & Partners