Generative AI and copyright conundrum

The Hindu

23-04-2025

The celebrated legal academic and copyright law scholar Paul Goldstein, in his copyright treatise, argues that there are two types of copyright law experts: 'the copyright optimist' and the 'copyright pessimist'. The optimist views the copyright's glass as half full and argues for additional rights to fill the cup to its brim. The pessimist views it as half empty and argues for satiating the copyright owner/creator with sufficient rights to protect the creative interest, while also arguing for a larger public domain that inspires and facilitates the future of creativity.
The copyright law was created to equip publishers with rights that evolved with the creation of the printing press. The copyright owner is bestowed with ownership over the work for a limited number of years. This is done with the intention of helping the owner to reap the rewards that are offered by the market. Once the ownership period ends, the work enters the public domain, whereby it becomes common for all to use so that the creative sea never dries.
This premise had been challenged by disruptive technologies several times in the past, and the copyright law has adapted to the challenges posed by technology. The law provides certain exceptions to infringement such as private use and non-commercial use. This is again an area where the copyright optimist and the pessimist differ. The pessimist argues for user right and the optimist opposes it.
For instance, when the photocopying machine was invented, it was a common fear among the creative community that copying would become rampant and the law-given monopoly would no longer remain meaningful. Similarly, when the Internet boom occurred, it was again felt that the creative works were open in the virtual space that enabled people to have access to copyrighted work by bypassing the legal route. In all these instances, the status quo of the law was maintained, and it did not challenge the fundamental norms of the copyright law. It pertained only to tailoring the copyright law's exceptions
New creator
In the recent past, this notion has been categorically challenged by generative AI. The previous technological development(s) only enabled third parties to have access to the copyrighted content, but generative AI models threaten to replace the human author, as AI dons the creative hat now.
Open AI, a forerunner in the generative AI field, faces legal challenge for falling foul of the copyright law. It has especially been accused of training the AI platform on copyrighted content without permission. This is a novel argument in copyright law, as contended by experts and generative AI companies that copyright law should operate on the 'creation' part and not on the 'means to creation'. If the means to creation is stopped, it would dry the creative pool.
Further, these models are educated on a vast trove of materials available on the Internet, and it would be difficult to 'pinpoint the exact occurrence of infringement', as the platform learns the nuances from the existing creation and creates based on prompts given by the user. This is like any creative work made by humans — even Mozart and Beethoven heard classical piano before they began to play it (this argument's legal validity is yet to be decided by courts across the world including India).
Last month, the Studio Ghibli trend began to surface on the Internet, with Internet users engaging ChatGPT to create their images in the animation style of the studio based in Japan.
The users might have had an amusing experience in styling their image in Ghibli style for free, but it sparks a serious copyright concern. The platform is able to generate images on request in a particular style because it has been trained on that style by its creators, and hence it is a violation of the copyright law, if done without permission.
Even if the Ghibli style images are created for non-commercial private use of the users, it is still a derivative work of the copyright owner, a work based on the copyrighted works of Studio Ghibli. This instance nullifies the previous argument of the generative AI companies before the courts that it is almost impossible to trace the exact work of infringement.
The platform is virtually trained on all materials from the Internet, and the generative AI work is nothing but a complex mixture of its learnings and is not particularly similar to one style. This argument may not be true in the current instance, as the AI is able to generate images in a particular style, in this instance, Studio Ghibli. It nudges one to think about the specific training that the AI has received from these works.
It also raises another concern that the AI begins to act as a potential market substitute for Studio Ghibli's works, leading to a possible market loss for it. Tomorrow, for a commercial use in a movie or in a newspaper article, the AI generated work could be preferred to the studio's original work, which is a lost economic opportunity for the company. This trend could dilute and effectively run contrary to the norms and theories of the copyright law. It runs the risk of throwing the cup with the water.
(The author is an Assistant Professor of Law at KIIT School of Law; views are personal)