I. ABSTRACT
This article explores the legal implications of AI-generated content, focusing on music and creative works. It delves into the functioning of Generative AI, like GANs, and their reliance on real artists' input. The global perspective covers initiatives in China, the European Union, and the United States, emphasizing the need for regulation. Legal challenges, such as copyright disputes and the limitations of AI authorship, are highlighted through cases like MuseNet and a revoked copyright for an AI-generated graphic novel. The absence of clear laws, both globally and in India, poses challenges in addressing copyright, fair use, and monetization concerns in the evolving landscape of AI-generated content.
II. INTRODUCTION
Artificial Intelligence (AI) have the means to generate content like images, videos, text, or any audio with zero human intervention, a breakthrough achieved in the AI world. World Economic Forum explained that a category of AI algorithms that generate new outputs based on the data they have been trained on, are Generative AI.
Generative AI uses deep learning to generate outputs. Deep learning trains AI through iteration of original data until acceptable generative data is created by the AI. Deep learning models like Generative Adversarial Networks (GANs) and variational autoencoders (VAE) help to achieve this same goal. Generative models find deviations and anomalies in the generated content. GANs consist of two neural networks i.e. Generator and a Discriminator that work simultaneously to comply with AI with the desired result.[ii] The generator thrives to produce music or audio similar to the original audio (input) and the discriminator keeps on rejecting the generator’s work until the generator starts to generate audio that resembles to original audio. The discriminator carries out its work without any human intervention.
Inputs used by AI for their learning process are the songs created by real artists. These songs are mostly freely available to the public. The Discriminator uses these songs to analyse the audio/song/music/voice produced by AI and it rejects all the music produced by the AI if that specific music does not match the prescribed music. The work of AI derives from the input it has received. Only a small amount of data is approved by the discriminator out of all the generated work. In April of 2023, ghostwriter977 posted the song “Heart on my sleeve”, the song had the voices of Drake and weeknd but the artists did not sing it rather their voices were mimicked by an AI.[iii] It may be possible by uploading the voices of respective artists to deep learning models. Songs created by AI are derivative songs because AI remains limited to its provided input.
A vast amount of data is collected for the purpose of training AI. The sources that are used to collect the data include web scraping and using pre-existing datasets which are freely available online.[iv] No credits are given to the rightful owners from whose data AI gets trained and compensating for the same is a distant talk. Even the consent of the owners is not taken for carrying out this process on their products. The output cannot be generated out of nothing by AI but they are derivative works of the real artists. In other words, the work generated by AI are rearrangement of pre-existent work and not something inexperienced but remains limited to their respective inputs.
III. GLOBAL SCENARIO
Developments in AI have not been limited to any one or two countries. It is a matter which includes a majority of countries. Developed and developing countries are in a race to understand this new technology and each country wants themselves to stay ahead in this AI game.
Countries are trying to understand the effects of AI on human society and formulate adequate laws to suitably regulate Artificial Intelligence in their respective jurisdiction, yet few to no countries have been successful. India has not updated its statutory to be inclusive of AI. In the Copyright Act, 1957 the word ‘author’ has been defined, where it is mentioned that “in relation to any literary, dramatic, musical or artistic work which is computer-generated, the person who causes the work to be created” is author.[v]
Countries across the globe have been formulating their plan of action regarding AI in future. the Chinese Government released the 'Next Generation AI Development Plan' in September 2017. It was issued by the State Council of China, emphasizing the importance of AI in driving economic and social development. A three-step strategy for AI development, culminating in 2030 with becoming the world's leading centre for AI innovation.[vi] China's focus on core technologies such as quantum communication and quantum computing, brain science, and deep underwater space exploration was highlighted in the document.
European Union will near future implementing their newly formulated AI Act. This Act has taken a “risk-based approach,” level of risk from artificial intelligence.[vii] The risks are divided into four parts which are unacceptable risk, high risk, generative AI, and limited risk.[viii] The AI Act of the European Union will be the first of its kind.
United States is ready with a blueprint for an AI Bill of Rights. The blueprint talks about data privacy, protection, and safety from unsafe and ineffective systems.[ix] It also underscores protection from algorithmic discrimination. Such algorithmic discrimination may violate the legal protections of an Individual living in a human society.
According to NITI Aayog, AI has the potential to add 1 trillion INR to the economy by 2035 in India.[x] India's AI strategy, known as AI for All, seeks to harness the advantages of AI for both economic growth and social development, emphasizing 'inclusive growth.' The primary focus is on empowering citizens to secure higher-quality employment opportunities.
G20 in 2023, under the presidency of India, has recognised that AI technologies, if not used responsibly, can have detrimental effects on society and intellectual property rights. Unauthorized use of AI models can lead to IP infringement, as proprietary algorithms and data are misappropriated. Furthermore, AI-powered systems may inadvertently discriminate against certain groups, leading to ethical and legal challenges.[xi]
IV. RECENT JUDICIAL INTERPRETATIONS
The first software to create music dates back to 1997, and since then there have been numerous advancements in the music industry, MuseNet is one such example which was developed by Sam Altman-led OpenAI, it is a deep neural network that can generate 4-minute musical compositions with 10 different instruments, and can combine styles from country to Mozart to the Beatles. MuseNet was not initially programmed with an understanding of music, but instead discovered patterns of harmony, rhythm, and style by learning to predict the next token in hundreds of thousands of MIDI files.[xii] But there are limitations to its capabilities, one of them being that the AI finds it difficult to pair unusual combinations of instruments, If the training data did not include a significant number of examples with unusual instrument pairings, the model might struggle to generate coherent and aesthetically pleasing compositions with those combinations.[xiii]
When we talk about AI-generated music content, there are two major issues which need to be dealt with. Firstly, can the use of intermediate material be termed as fair use and secondly, can the content generated by AI technology be copyrighted?
In the case of MuseNet, the OpenAI has collected training data from many different sources. ClassicalArchives and BitMidi donated a large collection of MIDI files for this project, and several collections online, including jazz, pop, African, Indian, and Arabic styles have also been used.[xiv] In similar fashion other AI generator tools also use the same methodology to train their AI, the matter of contention in this context is that a lot of the data used for training AI are copyrighted material, recently a class action suit was filed by two authors, Mona Awad and Paul Tremblay in the San Francisco federal court claiming that the organisation breached copyright law by “training” its model on novels without the permission of authors.[xv]
The black box effect of machine learning algorithms obstructs the transparency and explainability of the decision-making process.[xvi] The number of agents potentially involved adds to the complexity of the task and defining the liability and ownership becomes even more difficult for the regulating authority. This is the first lawsuit against ChatGPT that concerns copyright, this lawsuit is pending before the court as of now, but the observations made by the court will be detrimental in establishing a legal framework and the fair use doctrine in relation to the AI-generated content.[xvii]
Subsequently, one can argue that EDM or Electronic Dance Music is also generated using computers, While EDM is audio produced by computers, the ultimate creativity comes from a human who creates original music using the technology. On the other hand, an Al capable of generating music is likely to be trained on copyrighted material. This assertion comes from the fact that most Al-based music generation platforms can take commands such as “write music in Kanye West style”. So, in order to understand Kanye West style of singing, it needs to be trained on that data.[xviii]
Furthermore, in a similar case in September 2022, Ms. Kashtanova was granted a copyright registration for the partially AI-generated graphic novel, Zarya of the Dawn, by the USCO. But the copyright was later revoked by the USCO. Under US law, only a human can be an author, and the USCO’s Copyright Compendium explicitly states this. The USCO concluded that Ms. Kashtanova is the author of the Work’s text as well as the selection, coordination, and arrangement of the Work’s written and visual elements and that this is protected by copyright.[xix]
However, it observed that the images in the Work that were generated by AI are not the product of human authorship. This decision is significant because this would imply that the parts generated by an artist using AI in his musical work might not be protected under copyright, this generates various legal implications for other forms of creative works, including music, where AI technologies are increasingly involved in the generation of content.
When we talk about the protection of such content, there is no common consensus or explicit law regulating it. Although in the case of Urantia Foundation v. Maaherra[xx] regarding the copyright of a text supposedly authored by “celestial beings”. The Ninth Circuit Court observed that there needs to be a distinguishable element of human creativity involved in the creation, of a book to be copyrightable.[xxi] The court in the present case emphasised the requirement of human authorship for producing copyrightable content.
As long as the monetisation of such musical content is concerned, popular streaming platforms like YouTube have been in the process of developing updated policies regarding the uploading and monetisation of AI-generated music. Since there are no specific laws regulating the issue, the streaming platforms are finding it difficult to strike a balance between rights protection with creator-led innovation, including figuring out ways to monetize AI-generated content.[xxii] Although there are some companies like SOUNDRAW which is basically an AI music generator, that allows creators to make songs based on the selection of their mood, length etc.[xxiii] It works on a limited license model where you can monetise with YouTube or other streaming platforms but you cannot register or copyright claim the music generated by SOUNDRAW itself.
V. INDIAN SCENARIO
Although when we talk about the Indian context, there is no explicit judicial or legislative clarification on AI generative content, however, the application that identified AI (RAGHAV) as the sole author of an artwork was denied by the Indian copyright office. The Copyright Office then approved the application and listed AI (RAGHAV) and a natural person as co-authors; although they soon revoked the copyright. The appeal is pending a ruling as the applicant has contested the withdrawal notification.[xxiv]
On similar lines, the issue of voice cloning has increased significantly, In the Indian context, recently the famous actor Amitabh Bachan filed a suit for the protection of personality rights and secured an ad-interim ex-parte order for the New Delhi High Court. He has relied on his rights under the Copyright Act 1957, which grants the right to authors and performers to claim credit and authorship of their work under Sections 38 (performer’s right), 38A (exclusive rights of performers) and 38B (moral rights of performers).[xxv] The issue of voice cloning has increased significantly in recent times since AI generative tools are readily available on the internet and can clone any voice with a sample of just a few seconds.
Subsequently, when we discuss the use of training data in AI technology, Section 52 of the Indian Copyright Act deals with defences against copyright infringement.[xxvi] There are precedents like the case of Syndicate of the Press of the University of Cambridge and Anr v. B D Bhandari L[xxvii] where the court has observed that the purpose served by the subsequent work should be substantially different from the purpose served by the prior work.[xxviii] To be called transformative, the subsequent work must be different in character, it must not be a mere substitute.[xxix]
In India, the judicial and legislative opinions on AI-generated are yet to be developed. These precedents although help in developing a nascent understanding, deal with the issue of copyrightability of AI-generated content. A separate extensive judicial review is required to develop a legal framework for the same.
VI. CONCLUSION
Generative AI has the potential to significantly transform various aspects of the world across multiple domains and Generative AI has shown a substantial impact on the music industry. AI-driven music composition, synthesis, and even personalized listening experiences redefine artistic boundaries. This has led to a significant transformation of this industry. However, this revolution necessitates careful governance. AI is a human production that should neither be idolized nor demonized, but rather evaluated with a critical spirit, both in its benefits and risks for the preservation of humanity.[xxx] The European Union is presently developing enforceable legislation to regulate AI systems, underscoring the need for a global consensus on AI governance. Reaching global consensus is a complex task since different countries adopt different approaches. However, technologies like AI transcend national borders. A uniform set of rules can prevent discrepancies in AI regulation, ensuring a level playing field for all stakeholders.
[i] The authors are the students at Rajiv Gandhi National University of Law, Patiala.
[ii] Kinza Yasar & Sarah Lewis, Generative Adversarial Network (GAN), TechTarget (Mar. 9, 2023), https://www.techtarget.com/searchenterpriseai/definition/generative-adversarial-network-GAN.
[iii] Joe Coscarelli, An A.I. Hit of Fake ‘Drake’ and ‘The Weeknd’ Rattles the Music World, THE NEW YORK TIMES (Apr. 19, 2023), https://www.nytimes.com/2023/04/19/arts/music/ai-drake-the-weeknd-fake.html.
[iv] Sam O’Brien, AI Art: Why Some Artists Are Furious About AI-Produced Art, Ieee Computer Society (Jan. 15, 2024), https://www.computer.org/publications/tech-news/trends/artists-mad-at-ai
[v] Copyright Act, 1957, No. 14, Acts of Parliament, 1957 (India).
[vi] Eleanor Bird, Jasmin Fox-Skelly, Nicola Jenner, Ruth Larbey, Emma Weitkamp and Alan Winfield, National and International Strategies on AI, The ethics of artificial intelligence: Issues and initiatives (Jan. 15, 2024), https://www.europarl.europa.eu/RegData/etudes/STUD/2020/634452/EPRS_STU(2020)634452_EN.pdf.
[vii] Jennifer G. Betts, Simon J. McMenemy, and Danielle Ochs, EU Agrees on Groundbreaking AI Act: A Potential Model for AI Regulation? Ogletree (Jan. 15, 2024) https://ogletree.com/insights-resources/blog-posts/eu-agrees-on-groundbreaking-ai-act-a-potential-model-for-ai regulation/#:~:text=AI%20systems%20deemed%20to%20pose,emotion%20recognition%20in%20the%20work.
[viii] Artificial Intelligence Act, European Union, (March 13 2024) https://www.europarl.europa.eu/doceo/document/TA-9-2023-0236_EN.pdf
[ix] Blueprint for an AI Bill of Rights, https://www.whitehouse.gov/ostp/ai-bill-of-rights/.
[x] Responsible AI for All: Adopting the framework – A use case approach on Facial Recognition Technology, https://www.niti.gov.in/sites/default/files/202211/Ai_for_All_2022_02112022_0.pdf
[xi] India’s G20 Presidency: A Strategic Vision for Intellectual Property, Trade, and Responsible Artificial Intelligence, The IP Press, https://www.theippress.com/2023/09/30/indias-g20-presidency-a-strategic-vision-for-intellectual-property-trade-and-responsible-artificial-intelligence/
[xii] MuseNet, OpenAI (Jan. 14, 2024, 6:18 PM), https://openai.com/research/musenet.
[xiii] 13, Sunray Eric, TRAIN IN VAIN: A THEORETICAL ASSESSMENT OF INTERMEDIATE COPYING AND FAIR USE IN MACHINE AI MUSIC GENERATOR TRAINING, “IP Brief”, 1-32 (2021).
[xiv] Supra Note 11
[xv] Ella Creamer, Authors file a lawsuit against OpenAI for unlawfully ‘ingesting’ their books, The Guardian (July 5, 2023), https://www.theguardian.com/books/2023/jul/05/authors-file-a-lawsuit-against-openai-for-unlawfully-ingesting-their-books
[xvi] Custers Bart, LAW AND ARTIFICIAL INTELLIGENCE XXXX (Eduard Fosch-Villaronga eds., 2022), https://doi.org/10.1007/978-94-6265-523-2
[xvii] OpenAI faces lawsuit filed by US-based authors Mona Awad, Paul Tremblay, The Economic Times, https://economictimes.indiatimes.com/news/international/us/openai-faces-lawsuit-filed-by-us-based-authors-mona-awad-paul-tremblay/articleshow/101677923.cms?from=mdr
[xviii] As Grammys slaps ban on AI music, copyright question makes waves; Discord caught in row too, HT Tech, https://tech.hindustantimes.com/tech/news/as-grammys-slaps-ban-on-ai-music-copyright-question-makes-waves-discord-caught-in-row-too-71687940191446.html
[xix] Why a US legal ruling over an AI-generated cartoon has major implications for music’s future. - Music Business Worldwide, Music Business Worldwide, https://www.musicbusinessworldwide.com/why-a-legal-ruling-over-an-ai-generated-cartoon-has-major-implications-for-musics-future/
[xx] Urantia Foundation v. Maaherra,895 F. Supp. 1347.
[xxi] Urantia Foundation v. Maaherra, U.S. District Court for the District of Arizona, Feb. 10, 1995, 95 F. Supp. 1347, at XXXX (United States).
[xxii] Emilia David, YouTube is figuring out its AI strategy by working with music labels, The Verge (Aug. 21, 2023), https://www.theverge.com/2023/8/21/23840026/youtube-ai-music-copyright-monetization-universal.
[xxiii] SOUNDRAW, AI Music Generator - SOUNDRAW, https://soundraw.io/license (last visited Jan. 14, 2024).
[xxiv] The Copyright Conundrum: Exploring The Impact Of Generative Ai On Intellectual Property Rights, GSL Chambers, https://gslc.in/the-copyright-conundrum-exploring-the-impact-of-generative-ai-on-intellectual-property-rights/
[xxv] Big B, 'Computer ji, Lock kiya jaaye' — what Amitabh Bachchan’s suit for personality rights means, The Print, https://theprint.in/india/big-b-computer-ji-lock-kiya-jaaye-what-amitabh-bachchans-suit-for-personality-rights-means/1247185/
[xxvi] Doctrine of Fair Dealing in Indian Copyright Law – SURANA & SURANA, SURANA & SURANA – International Attorneys, https://suranaandsurana.com/2022/09/02/doctrine-of-fair-dealing-in-indian-copyright-law/#_ednref11
[xxvii] University of Cambridge v. B.D. Bhandari, 2011 SCC OnLine Del 3215.
[xxviii] Dipak Rao and Sana Singh, Acceptable Use of Copyrighted Material, Singhania & Partners, https://singhania.in/blog/acceptable-use-of-copyrighted-material (last visited Jan. 14, 2024).
[xxix] Abhinav Shrivastava and Radhika Jalan, The Copyright Conundrum: Exploring The Impact Of Generative AI On Intellectual Property Rights, https://gslc.in/the-copyright-conundrum-exploring-the-impact-of-generative-ai-on-intellectual-property-rights/
[xxx] Supra Note 15
Commenti