A Look at Generative AI in Terms of Intellectual Property Rights
**Aafreen Abdul Sathar
Introduction
Intellectual Property Rights (IPR) serve as a shield for inventors to be able to utilise and market their products exclusively in the form of various laws such as copyrights, patents, trademarks, and more. However, the growing field of Artificial Intelligence (AI) has sparked debates regarding the threat generative AI poses to the rights of human creators.
The World Intellectual Property Organization (WIPO) has defined the works of generative AI as the “generation of an output by AI without human intervention.” Generative AI has been accused of violating the IPR of human creators such as writers, artists, and musicians as its algorithm trains itself on their material, causing many to consider AI-generated matter ‘unoriginal.’ Furthermore, the nature of its creation has led to debates regarding the legal eligibility of AI-generated works for IPR protection and whether the IP rights of any AI-generated output belong to the user of the program responsible for the input, the owner of the program or the AI programme itself.
AI and The Process of Creation: How Generative AI Works
Generative AI uses neural network techniques such as Generative Adversarial Networks (GAN), Variational Auto-Encoders (VAE) or Transformers to process large sets of pre-existing data from the internet based on inputs such as prompts or keywords and create new outputs such as art, music, essays etc. by finding and analysing the features, patterns and styles of the data. The process by which generative AI creates its output has led to controversies and ethical and legal issues regarding copyright, which will be discussed further below.
Generative AI and its Battles with Copyright
When we consider the works in which copyright can subsist according to Section 13(a) of the Indian Copyright Act, 1957, which refers to copyright existing on “original literary, dramatic, musical and artistic works,” it raises questions regarding eligibility of copyright on AI generated outputs for no proper definition has been provided for the key term “original.” Due to generative AI using pre-existing content on the internet and generating material largely based on such content, it is debatable whether its output can truly be considered “original” and hence eligible for protection under copyright.
Two doctrines are commonly used to judge whether a work can be considered original: The Sweat of the Brow Doctrine and the Modicum of Creativity Doctrine. The Sweat of the Brow Doctrine considers a material to be original if the author had used skill and labour in its creation and it originated from him. Later, courts resorted to using the Modicum of Creativity Doctrine instead, ruling that for a work to be considered original, it had to have a minimum amount of ‘creative originality.’ Although AI-generated outputs do not meet the requirements of the first doctrine, it is debatable as to whether such material can really be considered to be ‘creative’ under the second doctrine as AI does not have truly novel ideas, creating an output based on already existing material from the internet and at many times, even replicating them.
Moreover, AI-generated outputs face a huge backlash from many creators due to their violation of IPR of human creators, using copyrighted material in the training and output creation of AI without permission from the original authors. Companies like GitHub and its parent company Microsoft and OpenAI are facing a class action lawsuit for “software piracy on an unprecedented scale” in the AI-generated code programming done by GitHub Copilot.
Furthermore, companies like Stability AI and Midjourney are being sued for their subsidiary companies, Stable Diffusion and Midjourney, that have violated the IPR of “millions of artists” by using about five billion images “without the consent of the original artists” for training their AI and using it to generate art. Stability AI is also being sued by Getty Images, the latter accusing the former of having “unlawfully copied and processed millions of images protected by copyright.” In various instances, AI-generating platforms like DALL-E 2 and Stable Diffusion have replicated and reused features from the copyrighted art used to train them. Moreover, it has been revealed that AI-generated content by companies like CNET has committed plagiarism on a mass scale by copying and paraphrasing work from many human writers.
AI has even developed enough to generate songs and covers emulating the voices of singers such as Jay-Z, Rihanna, The Weeknd and most popularly, Drake, fuelling the controversy around AI. After TikToker Ghostwriter977 used generative AI to create a new viral song using the voices of Drake and The Weeknd titled ‘Heart On My Sleeve,’ Universal Music Group has asked for the song to be removed from all the major streaming platforms as it ‘represents both a breach of our agreements and a violation of copyright law’. The record label believes that the use of generative AI in such a way harms their artists and infringes upon their IPR. AI is able to use the artist’s voice in any manner, which has caused concern over what these voice replications mean for the future of artists.
Generative AI also raises questions about the eligibility of works generated by AI liable for protection under IPR, with the most notable case being that of the painting ‘Suryat,’ inspired by Van Gogh’s Starry Night, which listed its co-authors as Robust Artificial Intelligent Graphics and Art Visualizer (RAGHAV) and Ankit Sahni in 2021, after the Indian Copyright Office initially denied the first application listing RAGHAV as the sole author. Eventually, even the co-authorship of RAGHAV was withdrawn as officials believed that in reference to Section 2(d) of the Indian Copyright Act, 1957, clauses (iii) and (vi) stated that an ‘author’ meant “in relation to an artistic work other than a photograph, the artist” and “in relation to any literary, dramatic, musical or artistic work which is computer-generated, the person who causes the work to be created” and hence, the AI could not be considered the co-author. Sahni, however, argues that there is no proper definition for the term ‘person’ in the Indian Copyright Act, which is a gap in the recognition of IPR of works generated by AI in India. On the other hand, Canada has chosen to grant ‘Suryat’ copyrights, listing Sahni and RAGHAV as co-authors. Cases like that of RAGHAV’s have led to debates regarding IPR for the outputs of generative AI regarding any ‘original’ material it may create as globally, there are few provisions regarding the authorship of AI and whether generative AI can receive IPR for its outputs or if not, whether the rights it will go to its user or owner.
The Outlooks: Generative AI Users vs Human Creators
Many sections of the general population view generative AI in a positive light, one deleted tweet even stating that AI “democratises creative expression.” Now people can use AI in order to make art, music, essays, codes, etc. and create a variety of outputs within seconds. However, human creators are of an entirely different opinion.
A tweet by Australian artist Zed Edge has gone viral on the platform for its view on generative AI: “When referencing art, humans use them as inspirations [and] AI uses them as samples. Without references, humans can visualise ideas, but AI cannot. Both benefit from references, but AI depends on them. AI doesn’t ‘learn’ from art, it manipulates art beyond recognition as a deception to appear new. Without consent or compensation to the original artists, AI art is theft.” Artists all over the internet seem to agree that the unchecked development of generative AI is leading to the collapse of their creative careers, and it has led to artists posting pieces with the line “NO TO AI GENERATED IMAGES” on social media platforms, commonly with the hashtag #SupportHumanArtists.
When one looks at cases like the Colorado State Fair 2022, where Jason Allen won first place in the digital arts category of the fine arts competition by inputting various prompts into MidJourney to generate a painting he called ‘Théâtre D’opéra Spatial,’ there is no wonder that all sorts of artists are upset about the development of generative AI. “We’re watching the death of artistry unfold right before our eyes,” tweets OmniMorpho, a sentiment shared by many artists who believe that not only are their jobs in danger but their works are even being used without their consent to train and develop the style of their competition. With mangled bits and pieces of a watermark or signature from the art piece being used to train the AI showing up in the generated output, many believe AI is only mixing, rearranging and stealing the content of human creators and infringing upon their IPR.
Many musicians and record labels are also against AI-generated covers and songs that use artists’ voices, with Universal Music Group stating that “The training of generative AI using our artists’ music (which represents both a breach of our agreements and a violation of copyright law) as well as the availability of infringing content created with generative AI on DSPs, begs the question as to which side of history all stakeholders in the music ecosystem want to be on: the side of artists, fans and human creative expression, or on the side of deep fakes, fraud and denying artists their due compensation.” Writers, photographers and other content creators are also opposed to generative AI that plagiarises their work and gains profit from it while reducing their opportunities.
Generative AI in the EU and Other Jurisdictions
Following the application of the Sweat of the Brow and the Modicum of Creativity doctrines in determining the originality of a work, the European Court of Justice, in the Infopaq case, concerning the interpretation of Directive 2001/29/EC or the directive on harmonisation of certain aspects of copyright, held that copyright under the directive was only applicable in cases where the subject matter was “original in the sense that it is its author’s own intellectual creation.” Subsequent judgements by the Court of Justice have held that a work would be considered an author’s own intellectual creation when the author had successfully utilised and expressed creative abilities in the production of the work by making free and creative choices which shall “stamp the work created with their personal touch.”
Furthermore, the Prague Municipal Court noted that an image cannot, in principle, be protected by copyright if it was generated by AI because it was not the product of the creative endeavours of a natural person or an author and, hence, was not eligible for copyright protection. However, the Beijing Internet Court has taken an opposite view, noting that as a generative AI model had no free will and was not recognised as a legal subject, the process of creation through generative AI “is a process of man using tools to create, that is, it is man who does intellectual investment throughout the creation process, not the AI model.” Ultimately, it held that as long as the AI-generated images “reflect people’s original intellectual investment,” the works were protected by copyright law, and the user who input prompts for the specific generation would be recognised as its author.
The United States Copyright Office, relying on multiple court judgements, released its registration guidelines on AI-generated works, clarifying that copyright only subsists on the product of human creativity and not on AI-generated content. In some cases, the human-authored aspects of the generated work or the extent of creative control of the author could be protected by copyright. An example of the same is when the US Copyright Office withdrew the initial complete copyright it gave ‘Zarya of the Dawn,’ a graphic novel generated by the AI MidJourney using prompts provided by author Kris Kashtanova upon realising the origins of its images, but provided the copyright to the arrangement of text and pictures and the novel’s storyline but not the images which were not the “product of human authorship.”
In 2024, the European Parliament took a large stride forward in the debate between generative AI and human creators by laying down Regulation 2024/1689, or the Artificial Intelligence Act, regarding the rules on artificial intelligence and the protection of IPR of human artists. Clause 105 of the Act, understanding the vast amount of data required and utilised for the training and output of generative AI, including material protected by copyright, has held that the use of such copyright-protected content in generative AI shall require the authorisation of the copyright holder concerned. It has added that “where the rights to opt-out has been expressly reserved in an appropriate manner, providers of general-purpose AI models need to obtain an authorisation from rightsholders if they want to carry out text and data mining over such works.” Furthermore, the Act holds that regardless of the jurisdiction where the training of generative AI models takes place, the providers are liable to ensure policies that comply with Union law on copyright and require them to publicly provide a detailed summary of the content used for training its AI model. The regulation acts as a landmark step in protecting and preventing infringement of human artists’ IP rights by generative AI while also allowing for the evolution of generative AI.
Conclusion
The nature of the production of generative AI and its rapid rampant growth not only threatens the livelihood of human creators but also infringes upon their IPR. The law, particularly in India, has yet to catch up with the recent developments in AI, and there is a global requirement for more laws to regulate the developments of Generative AI in such a manner that it allows both human creators and AI to co-exist in a peaceful, safe manner that does not threaten anybody’s right to creative expression.
Human creators have a right to safeguard their outputs and direct their use, reproduction, distribution, translation, adaptation, etc., unless otherwise needed for public welfare, and lawmakers have the responsibility to protect that right. AI-generated outputs are the result of training the AI using pre-existing datasets drawn from the internet and, in doing so, often uses copyrighted material such as essays, codes, art and so on without permission. For the training of such AI, the datasets must be carefully curated with works belonging to the public domain and works provided for training with consent from the authors in order to prevent plagiarism and copyright infringement. Moreover, the debates regarding the legal eligibility for IPR of works created by generative AI must reach a proper conclusion and enact legal provisions in order to allow for further progressive development that can benefit humankind and AI.
**Aafreen Abdul Sathar is a first-year law student pursuing a B.A L.L.B (Hons.) degree at Symbiosis Law School, Pune. Her interests include intellectual property rights, women’s rights and tort law. She is an active member of the Kautilya Society, a student-driven initiative at SLS Pune, in collaboration with Vidhi Centre for Legal Policy dedicated to policy research, legal discourse, and advocacy on contemporary socio-legal issues. Aafreen, in addition to her academic pursuits, has also engaged in social work with domestic violence shelters and schools for children with special needs. She is currently pursuing further research on intersectional feminism, adoption laws in India and the changing global landscape of environmental law.
Disclaimer: The views expressed in this blog do not necessarily align with the views of the Vidhi Centre for Legal Policy.