.jpg)
Author: Ayushi Shrivastava
Hitkarini Law College, Jabalpur
ABSTRACT:
In the age of artificial intelligence (AI), generative AI tools have become capable of producing a wide range of content, from texts and essays to visual art and music. However, this proliferation of AI generated works raises significant intellectual property concerns. But who owns the copyright for this content? Is it the developers of the AI tools, or the original creators of the material on which the AI was trained? Then Questions arise regarding the infringement of intellectual property rights by developers who collect and utilize content to train their AI models. 2023 saw unprecedented lawsuits arising from disputes over art creation using AI, with artists challenging the unattributed use of their work in AI created algorithms receive correctly. These ongoing class actions will play a central role in the future of such litigation. This article aims to delve deeper into the concept of copyright in different countries, the training process of AI systems, and the debate surrounding the originality of AI-generated content. It will also examine whether AI-generated works can be considered “original” or violative. Additionally, we will analyse major copyright lawsuits currently taking place in the AI landscape, providing insight into their impact and potential outcomes. This article will also discuss recent AI rulings that have profound implications for copyright protection of AI-generated works. By exploring these complex copyright challenges and examining real-life cases, this research contributes to a deeper understanding of the copyright landscape in the age of artificial intelligence.
Keywords: Copyright, Artificial intelligence, Authorship, Owner, Intellectual property right, etc.
INTRODUCTION:
Artificial intelligence (AI) has undeniably become an important and fascinating topic in many different legal fields, including liability, criminal law, legal technology and even law enforcement agricultural law. Therefore, the legal status of works created by artificial intelligence (AI) has caused much debate between national governments and researchers around the world. Copyright ownership of AI-generated content under copyright law and the need for a more flexible definition of copyright in the AI era are two major debated topics. Research into how intellectual property (IP) systems protect works created by AI is still in its infancy due to the complexity and rapid development of AI. Organizations such as the World Intellectual Property Organization are developing preliminary intellectual property policy considerations for AI in response to the need to address these emerging issues. National intellectual property offices have also sought to improve understanding of AI and intellectual property policies. However, as AI tools become more important to business workflows, so do concerns about possible intellectual property infringement claims and the ability to enforce content rights created by AI systems are emerging. Recent class action lawsuits, such as the one against Open-AI for alleged intellectual property infringement of the works used to train their AI models, demonstrate the importance of being aware of hazards and taking preventive measures. Important and decisive decisions in many areas were made in 2019 and 2020. Notably, in Europe, decisions relating to AI patent applications, especially those relating to DABUS and other AI-generated innovations, have been issued by both Offices European Patent (“EPO”) and UK Intellectual Property Office (“UKIPO”). These decisions have provided us with important information about how to assess who is the true inventor and how to apply patentability rules to advances made by AI. Inferences are made that DABUS and other AI systems should not be considered innovators and,
therefore, are not eligible for patent protection [1]. This article aims to explore the multifaceted aspects of copyright law in various countries, with a particular focus on the training process of AI systems and the ongoing debate over the originality of content AI-generated content. It will seriously consider the question of whether AI-generated works can be considered “original” or inherently infringing, taking into account different legal perspectives and frameworks. Additionally, the article will analyse the key copyright lawsuits currently shaping the AI landscape, providing valuable insights into [2] their impact and potential outcomes. He will also discuss the significant impact of recent cases currently pending as well as decided cases on the copyright protection of AI-produced works, highlighting the importance their importance in shaping the future of AI and copyright.
WHAT IS COPYRIGHT?
Copyright refers to the legal rights of the owner of intellectual property [3]. The legal concept of copyright gives authors sole control over the reproduction and use of their original works. This simply means that only the creator and those with their permission can copy or use the work. Copyright law ensures that artists retain ownership of their work and have control over how it is used and distributed. By allowing them to profit from their work for a limited period of time, it fosters innovation and
creativity while creating a framework that legally protects their rights. Although many nations have required originality as a condition for copyright protection, there are various definitions and requirements for what counts as original work [4].
Sweat of brow doctrine:
The theory is that authorship rights are acquired through careful work in creating the work. It doesn’t require much creativity or “originality”. Authors are entitled to these rights based on the effort and resources they expended in creating their work. For example, copyright protection may extend to certain works of authorship, such as telephone directories or databases, but this is not necessarily because they demonstrate creativity or original expression. This may be done because you have put a lot of effort, financial investment and commitment into organizing. Store data in a specific way. However, such compilations must be original to the author and must not be plagiarized. [5]
Modicum of creativity doctrine:
The definition of "originality" has evolved significantly in copyright law. The traditional "sweat of the brow" principle protected copyright based on an artist's work, talent, and investment, rather than originality. The standard that a work must exhibit a "minimum amount of creativity" to be considered unique was raised in Feist Publications Inc. v. Rural Telephone Service [6] case. The Supreme Court emphasized the value of "creative originality" and created a new standard for evaluating copyright protection based on basic creativity. Works must meet the requirements of the new standards of intellectual creativity and judgment. Copyright protection requires a minimum amount of creativity. Let’s take a look at how other countries approach assessing the originality of works in court, depending on the principles they use.
Position in English law (UK):
The sweat of brow doctrine was prominently defined in University of London Press v. University Tutorial Press [7], which was widely used in the United Kingdom.
The University of London Press v. University Tutorial Press [8] case is a notable legal precedent in which the Chancery Division of England clarified the standard of “originality” and is sometimes cited as an example of a “sweat of the brow” case. In this instance, the court ruled that the work did not have to be in an original or novel form to be covered by the Copyright Act.
However, it stipulates that the work must be original and not a copy from another source. The Court ruled that, because the exam questions in this case were written by the authors themselves, they were considered original works within the meaning of copyright regulations. The court emphasized that a plaintiff cannot be denied copyright simply because other examiners have raised similar issues. In the case of Walter v. Lane [9] and then at Ladbroke (Football) Ltd. V. William Hill (Soccer) Ltd [10].
In this case, the “sweat of the forehead” doctrine was used. In these decisions, courts have determined that the doctrine emphasizes the work, skill, and judgment required to create a work rather than the intelligence, accuracy, or literary value of the work. Other countries, including Canada, Australia and India, have also adopted the concept.
Position in the United States of America:
Since the late 17th century, American courts have recognized the value of authors' subjective and creative contributions. The literary and creative value of a work is also highlighted under the US copyright framework. In the United States, the criterion of originality requires "minimum creativity", as established in the Feist Publications decision [11], in which the US Supreme Court emphasized the need for creativity. Minimum creation is a necessary condition to obtain copyright protection and that applying judgment and work alone will not be sufficient to satisfy this requirement.
Position in India:
India has long advocated the “sweat of the brow” philosophy. The Supreme Court of India has accepted that copyright law does not prohibit borrowing useful parts of an original work while adding improvements, following the example of the English courts. Oriental Book Company v. Modak [12], in which the Supreme Court rejected the “sweat of the brow” theory and chose the American-style “minimum creativity” approach, is an important case in India on this subject. In this case, the idea of “minimal flavour.” Requires of “Creativity” was presented. Courts have held that to establish copyright, a work must demonstrate a specific degree of creativity.
HOW GAI (GENERATIVE ARTIFICIAL INTELLIGENCE) ARE TRAINED?
Machine learning is a key factor driving rapid advances in AI technology today. At its core, machine learning is the concept that computers can learn and improve performance by recognizing patterns without being explicitly programmed. As a subset of artificial intelligence, machine learning focuses on developing autonomous systems with the ability to learn and adapt without relying on precise instructions from human programmers. Computer programs can use machine learning to learn from a set of training data, make adjustments, and make future decisions based on the information learned independently or with guidance. The training data set is used by general artificial intelligence (GAI) systems like ChatGPT’s to generate new compositions while also including specific choices to influence their appearance. Although programmers can choose certain parameters, computer software generates its own work using neural networks, simulating cognitive processes observed in humans. This quality is the defining quality of this type of artificial intelligence.
In the case of language models like, machine learning involves leveraging existing works such as books, articles, and many other sources as part of the learning process. The model learns from this diverse range of input data to develop language comprehension and production. By leveraging this previously created work, the model gains information about language patterns, style, and context, allowing the model to produce responses that match the data on which it was trained. A common technique for obtaining input data for models like ChatGPT’s is web scraping.
This involves using specialized technologies to automatically extract data from websites, social networks and other Internet sources. ChatGPT's creator, OpenAI, takes input from a variety of collaborative and freely available sources, including material that may be in the public domain. However, there have been allegations regarding OpenAI's use of copyrighted materials obtained through web scraping to train its language model [13]. In response to these allegations, OpenAI asserts that its use of copyrighted works, including those owned by plaintiffs and other class members, falls within the limits of use legal. OpenAI
argued that its use of those materials was fair under fair use principles, exempting it from liability for copyright infringement.
AI AND COPYRIGHT:
1. The question revolves around the accountability of AI for copyright infringement.
2. Whether protection can be given to AI-generated work?
3. If protection is given to Ai generated work who will be the author the creator of AI, AI itself, or nobody?
Machine learning is a key factor driving rapid advances in AI technology today. At its core, machine learning is the concept that computers can gain knowledge and improve their performance by recognizing patterns without being explicitly programmed. As a subset of artificial intelligence, machine learning focuses on developing autonomous systems with the ability to learn and adapt without relying on precise instructions from human programmers.
Computer programs can use machine learning to learn from a set of training data, make adjustments, and make future decisions based on the information learned independently or with guidance. The training data set is used by general artificial intelligence (GAI) systems like ChatGPT to generate new compositions while also including specific choices to influence their appearance. Although programmers can choose certain parameters, computer software generates its own work using neural networks, simulating cognitive processes observed in humans. This quality is the defining quality of this type of artificial intelligence. In the case of language models like ChatGPT, machine learning involves leveraging existing works such as books, articles, and many other sources as part of the learning process. The model learns from this diverse range of input data to develop language comprehension and production.
By leveraging this previously created work, the model gains information about language patterns, style, and context, allowing the model to produce responses that match the data on which it was trained. A popular technique for obtaining input data for models like ChatGPT is web scraping. It involves employing specialized technologies to automate data extraction from websites, social media networks, and other internet sources.
ChatGPT's creator, OpenAI, gets its input data from a variety of collaborations and freely accessible sources, including pieces that might be in the public domain. However, allegations have arisen regarding OpenAI’s usage of copyrighted materials obtained through web scraping for training its language model [13]. In response to these allegations, OpenAI asserts that its utilization of copyrighted works, including those belonging to the plaintiffs and other class members, falls within the bounds of fair use. OpenAI argues that their use of such materials is justified under the principles of fair use, which exempts them from liability for copyright infringement.
When machines create who gets to own it?
1. Programmers as authors: This argument states that programmers should be the creators and legal owners of works created by AI. This tactic is used in countries such as Hong Kong, India, and the United Kingdom. For example, under UK copyright law, the author of a work is the person who takes the necessary precautions to create it.
2. AI itself as author: Some argue that AI should be given sole copyright if it creates original works on its own. However, this view faces challenges due to the legal recognition of AI as legal persons and the requirement that works resulting from human creativity and intelligence be copyrighted [ 14].
3. Prohibition of ownership or free use: Another perspective suggests that works generated by AI should be considered free without a copyright owner, similar to a Creative Commons license. This approach raises concerns for companies that have invested significant resources in the development of AI, as it could negatively impact their ability to innovate and derive economic benefits from the works produced.
Overall, approaches to the ownership of AI-generated works and copyrights vary across jurisdictions, raising important considerations regarding the role of programmers, AI systems, and broader implications for innovation and economic incentives.
THE RECENT’LANDMARK AI RULINGS
2019 and 2020, significant administrative and judicial AI decisions:
China:
The court issued an important decision in December 2019 upholding copyright protection for essays created by Tencent's AI system called Dream writer. Tencent has developed Dream writer, an intelligent writing support tool that generates documents. Tencent posted Dream writer’s financial analysis on the Tencent Securities website in August 2018[15].
However, Shanghai Yongxun Technology Company copied the same content and posted it on its website without obtaining formal permission from Tencent. Tencent sought legal protection for AI-generated content, and in response to this unauthorized copying, he took Yngxun to court for copyright infringement. By republishing the article without obtaining the necessary permissions, Yngxun is accused of infringing Tencent's copyright, which is the basis of the dispute. During the trial, two important questions were answered.
The court first considered whether the message demonstrated a basic level of originality. Second, the court found that the process of creating the new article revealed the Tencent team's personal preferences, judgment, and experience.
This article was created by Plaintiff's primary creative team through his four-step process: (1) Data services, (2) Trigger and write, (3) Intelligent verification, (4) Intelligent delivery. The court recognized that Tencent Development had control and direction over Dream writer's automated operations. Therefore, it is unfair and unfair to consider this automatic process to be the only and complete generation process performed only by Dream writer.
European Union:
In one notable case, Dr. Stephen Thaler, an expert in cutting-edge AI systems, applied for patents at the UKIPO [16] and the EPO [17]. In the filing, he claimed that an AI system called DABUS independently generated the patented idea and named him as its inventor.
However, the application was rejected by both his EPO and his UKIPO. The EPO said that both Article 81 of the European Patent Convention (EPC) and Article 19(1) of its Implementing Regulation, which
explicitly require the designation of a natural or legal person as the inventor, had been violated. They argued that the legal structure of the EPC limits inventors to humans or legal entities, as machines lack the legal personality and privileges associated with inventors.
Similarly, the UKIPO concluded that UK patent law recognizes only human inventors as true inventors, and that AI systems are not considered human and have no rights. According to the legislative history of their respective patent laws, both the EPO and UKIPO have based their decisions on the principle that the inventor must be a natural or legal person.
The UK Supreme Court further upheld the UKIPO's decision, reaffirming established precedent prohibiting corporations and other non-human entities from being recognized as inventors. A patent application naming DABUS as the inventor was subsequently rejected [18].
United States:
Additionally, Dr. Thaler filed two patent applications with the USPTO, naming DABUS as the inventor of both DABUS inventions. However, the USPTO denied both requests for three main reasons. First, as several patent law examples make clear, the USPTO has held that the statute specifically requires that inventors be natural persons. “Inventor” is clearly defined in sections such as 35 U.S.C. 100(a) is the person or persons who made the invention and are the subject of the patent application.
In addition, the law also systematically uses pronouns associated with natural persons. Second, the USPTO ruled that existing case law does not support the recognition of AI systems or corporations as authors and that “only natural persons” can serve as inventors [19]. Furthermore, the USPTO emphasizes that invention requires a process of idea generation, including a clear and sustained conception of the invention. This important step requires uniquely human mental abilities, abilities that machines lack.
As a result, the AI system cannot meet the design requirements. These justifications led the USPTO to conclude that Dr. Thaler's patent application did not comply with 35 U.S.C. 115(a), resulting in the application being denied. The United States District Court for the Eastern District of Virginia subsequently received Dr. Thaler's complaint against the USPTO.
MAJOR AI COPYRIGHT INFRINGEMENT LAWSUITS (that are happening right now) Ai code generators:
Core issue: Ai code generators trained on open-source code – is there a copyright infringement?
A proposed class action lawsuit names Microsoft, its subsidiary GitHub, and business partner OpenAI, alleging that their development of an AI-powered coding assistant, GitHub Copilot, relied on “software copyright infringement.” software on an unprecedented scale. In this complaint, the companies in question are accused of systematically violating copyright rules. The case is still in its early stages, but it could have a huge impact on the AI sector, where companies make their fortunes by teaching software using copyrighted data.
AI art generators:
Core issue: Does AI-generated art violates copyright in the artwork that AI was trained on?
Stability AI, Mid journey and Deviant Art were sued by a group of artists including Sarah Andersen, Kelly McKernan and Karla Ortiz. These companies, Stable Diffusion, Mid journey and Dream Up, are in charge of producing AI works of art. The artists claim that by using five billion photos scraped from the web to train their AI algorithms without the permission of the original creators, these platforms are violating their rights. some artists. Butterick called the trial “a new step forward.” Towards fair AI & ethics for all” in a blog post announcing it [20].
Core issue: AI art generators trained on stock images – is there copyright infringement?
The creator of the widely used AI painting tool Stable Diffusion, Stability AI, has been sued by Getty Images for alleged copyright infringement. Getty Pictures said Stability AI improperly downloaded and processed millions of copyrighted images to train its algorithms in a statement provided to the Verge [21].
AI Language models:
Core issue: Language model trained on text found on the public internet-is there copyright infringement?
A lawsuit against OpenAI was recently filed in federal court in San Francisco by bestselling authors Mona Awad and Paul Tremblay. The authors have filed a proposed class action lawsuit, alleging that OpenAI trained the ChatGPT artificial intelligence chatbot using their copyrighted works without their permission. Authors Paul Tremblay and Mona Awad claim that ChatGPT was taught in part by “eating" their books against their will [22]. Sarah Silverman, along with authors Richard Kadrey and Christopher Golden, filed suit against OpenAI and Meta in a United States district court, alleging copyright infringement. The lawsuits claim that OpenAI's ChatGPT and Meta's LLaMA were trained on illegally obtained datasets, including their literary works from so-called "ghost library" sites like Library Genesis, Bibliotik, Z- Library and others. The plaintiffs claim that these books can be accessed in large numbers through torrent systems.
CONCLUSION
In summary, the outcome of the class action lawsuit against OpenAI for misappropriation of copyrighted material in LLM will have significant implications around the world. This tests an argument that has been debated in academia for many years. Creating his Marketplace, a database where writers, composers, and artists can use Smart, His Contracts to give his AI companies access to their works for a fee, addresses the issue of abuse. Here is one of his possible solutions proposed in. Although the court is not likely to specifically address issues of originality and copyright protection for his AI-generated work in this case, the court’s findings regarding copyright infringement may nevertheless influence subsequent decisions. In the age of AI, intellectual property protection must support creativity and innovation while paying artists fairly. Companies that create or use AI systems must take precautions to protect themselves from infringement lawsuits. This includes obtaining appropriate permissions, recognizing the author of the work, and regularly testing the AI system. Employers of AI solutions should read the terms and conditions carefully to ensure that AI tools do not produce unauthorized output. Policy and legislative advice from the Copyright Office and courts is needed to provide clarity and achieve a balance between copyright protection and innovation. Ultimately, negotiating the interface between AI and copyright requires rethinking our understanding of innovation and the legal incentives on offer.
REFERENCES
[1] Haochen Sun, Redesigning Copyright Protection in the Era of Artificial Intelligence, 107 IOWA L. REV. 1213,1215 (2022)
[2] Sylvia Polydor, Martyna Czapska & Karen Roberts, Chinese Dream writer Decision: A Dream Come True for AI Generated Works? BAKER Mc KENZIE: CONNECT ON TECH (Apr. 20, 2020),
[3] Will Kenton, Copyright Definition, Types, and How It Works, Investopedia, (15/07/2023, 20:22), https://www.investopedia.com/terms/c/copyright.asp
[4] The daily guardian, https://epaper.thedailyguardian.com/view/314/08-jul-2023/7
[5] Hailshree Saksena, DOCTRINE OF “SWEAT OF THE BROW”, Papers SSRN, (15/07,2023,20:30), http://papers.ssrn.com/sol3/papers.cfm?abstract_id=1398303
[6] Feist Publications Inc. v. Rural Telephone Service,499 U.S. 340, 342 (1991)
[7] University of London Press v. University Tutorial Press, England, Chancery Division. [1916] 2 Ch. 601 [8] Id
[9] Walter v. Lane, [1900] AC 539
[10] Ladbroke (Football) Ltd. V. William Hill (Football) Ltd., [1964] 1 WLR 273
[11] Feist Publications Inc. v. Rural Telephone Service,499 U.S. 340, 342 (1991)
[12] Eastern Book Company v. D.B. Modak, 2002 PTC 641
[13] Bridget Watson, A Mind of Its Own – Direct Infringement by Users of Artificial Intelligence Systems, 58 IDEA 65 (2017).
[14] Sainee Abhishek, Artificial Intelligence and Copyright Issues, Legal service India, (18/07/2023, 16:23),
https://www.legalserviceindia.com/legal/article-9895-artificial-intelligence-and-copyright-issues.
[15] e Bo Zhou, Artificial Intelligence and Copyright Protection-Judicial Practice in Chinese Couts, at 2 [hereinafter Dreamwriter ruling],wiPo, (16,07,23) https://www.wipo.int
[16] UK Intellectual Property Office
[17] European Patent Office
[18] Sainee Abhishek, Artificial Intelligence and Copyright Issues, Legal service India, (18/07/2023, 16:23),
https://www.legalserviceindia.com/legal/article-9895-artificial-intelligence-and-copyright-issues.
[19] James Vincent, AI art tools Stable Diffusion and Midjourney targeted with copyright lawsuit, Theberge (16/07/2023,16:10),
https://www.theverge.com/2023/1/16/23557098/generative-ai-art copyright-legal-lawsuit-stable-diffusion-midjourney-deviantart
[20] Haochen Sun, Redesigning Copyright Protection in the Era of Artificial Intelligence, 107 IOWA L. REV. 1213 (2022).
[21] James Vincent, AI art tools Stable Diffusion and Midjourney targeted with copyright lawsuit, the verge (16/07/2023,16:10),
https://www.theverge.com/2023/1/16/23557098/generative-ai-art copyright-legal-lawsuit-stable-diffusion-midjourney-deviantart
[22] Emily St. Martin, Bestselling authors Mona Awad and Paul Tremblay sue OpenAI over copyright infringement, https://www.latimes.com/entertainment-arts/books/story/2023-07-01/mona-awad-paul tremblay-sue-openai-claiming-copyright-infringement-chatgpt
Comments