The concerns of AI-generated content
With the development of artificial intelligence entering its next phase, this technology has sparked controversy about how to use it legally to simulate human behaviours and intelligent thoughts. Artificial Intelligence is also considered the ‘core technology that stands out in a new technological revolution, driving the enormous engine of economic and social development (Flew, 2021)’. McKinsey’s Artificial Intelligence Research Report indicates that “By 2025, the market share of AI will reach $127 billion… Artificial intelligence is driving a transformation in human society that is ten times faster than the Industrial Revolution (Bao, 2017).” Jerry Kaplan, a known Artificial Intelligence expert at Stanford University, states that the significant changes brought by AI technology are sufficient to redefine an era (“This Is Gonna Change Everything”: A Very Short History of Generative AI, n.d.).
Generative AI conducts machine learning techniques by allowing systems to learn and imitate users’ behaviours automatically based on collected data and information (Kahveci, 2023). For this reason, data is described as ‘the new oil’ – one of the most valuable resources – as a foundation for building machine learning and Large Language Models (Flew, 2021). At the same time, the emergence of Artificial Intelligence generators like ChatGPT, Midjourney, Stable Diffusion and DALL-E2 has significantly impacted people’s lives and careers. The text-to-image generators can implement artistic creation and the spiritual aesthetic of humans through computational power. However, these robots have also brought up several ethical and legal issues. The deployment of AI-generated content and its training data raises questions regarding authorship, intellectual property rights, and the authenticity of artistic expression that deserves investigation. It is imperative not only to focus on the progression and innovation of technology but also to elaborate upon its influence on human society and human creativities (“The Aesthetic Ethics of Midjourney Under the Development of Artificial Intelligence,” 2023). Furthermore, the potential for AI to replicate and even surpass human creativity that violates artistic authenticity simultaneously is worth scrutinizing. We must thoroughly examine how to set the ethical guidelines of AI toolkits and their potential risks to societal and human values. In the age of AI, it is ubiquitous that users have to confront plenty of ethical and legal challenges on the Internet that violate their privacy and digital rights, including issues about data surveillance, personal privacy, intellectual property rights and transparency in algorithms (“The Aesthetic Ethics of Midjourney Under the Development of Artificial Intelligence,” 2023). These dilemmas demand regulations to restrict the development of AI along the right path.
The lawsuits against generative AI websites
Since the beginning of 2023, significant legal cases have been raised against AIGC in the United States, drawing Congress’s attention to the necessity to enhance the boundaries of the AI landscape. In 2023, three artists, Sarah Andersen, Kelly McKernan, and Karla Ortiz, initiated legal action against Stability AI and Midjourney, the developers of AI art platforms Stable Diffusion and Midjourney, respectively, along with the artist portfolio platform DeviantArt, which has recently introduced its own AI art generator called DreamUp (Vincent, 2023). The artists argue that AI-generated robots are mainly developed upon the copyrighted creations of countless artists. It poses a threat to artists and photographers as millions of artworks and images in the training data are used without the explicit consent of the creators (Vincent, 2023). This lawsuit documentation includes a name list of approximately 4,700 artists whose artworks have been exploited by Midjourney “without the consent of the original artists (Nelson, 2024).” Lawyer Matthew Butterich launched the lawsuit in collaboration with the Joseph Saveri Law Firm, renowned in antitrust and class action cases. Butterick defines this case as a crucial step to balance fairness and ethics between AI and humans. He states that text-to-image tools like Midjourney “floods the market with an essentially unlimited number of infringing images will inflict permanent damage on the market for art and artists (Vincent, 2023).” Additionally, Reid Southern, an artist who is famous for his collaborations with Marvel and DC, also gained public attention by sharing screenshots from Midjourney’s Discord server, which are conversations involving Midjourney CEO David Holz. The screenshots prove that the program generated “significant amounts of MTC cards” during its testing stages, utilizing artists’ artworks as a basis for its algorithm (Nelson, 2024). Holz has conveyed the intention not to be drawn into issues regarding plagiarism and copyright. However, the current situation is that he and his company have eventually become involved in these intricate cases.
(Lang, 2023)
In a similar lawsuit with Getty Images, one of the biggest suppliers of stock images, photographs and videos, the company sued Stable Diffusion for using over 12 million stock photos to train the algorithms without obtaining permission or compensation (Brittain, 2023). The training dataset is usually collected from the public domain in such text-to-images toolkits. However, most of the contributions are still subject to copyright. This action violates the owners’ copyrights, where they post art pieces online, aiming to receive commissions and profits from them (Hayes, 2023).
Apart from these particular cases, wider attention has been increasingly attracted across different industries because there is a significant tendency for more companies to adopt AI art generators into their business as financially advantageous substitutes for hiring artists in person. Butterick and Saveri have also commenced similar cases against tech giants Microsoft, GitHub and OpenAI, where the AI programming tool CoPilot has the same issue that some of the codes are collected from the Internet. In the present era, AI has pervaded our daily lives and shaped them in profound ways. With its potential to provide creativity and efficiency in design and fine arts, it becomes imperative to critically assess the evolving relationship between human creators, AI-generated content, and artistic expressions in the digital age. This situation underscores the complex legal and ethical considerations that arise in machine learning technology, particularly regarding the utilization of data sourced from online platforms.
Legal challenges of AI-generated content in copyright law
It is expected that when writing academic articles, guidelines for plagiarism and proper citation sources are well-established. However, when it comes to the visual design and fine arts industry, the boundaries of plagiarism are often ambiguous. The article ‘The Problem with Plagiarism’ indicates that there is no borderline on how to distinguish between borrowing, referencing and plagiarism during postmodern design, where similar situations happen in the realm of AI technology. (The Problem With Plagiarism | Design Education Forum of SA, 2011). From the legal perspective, since the output of AIGC is produced based on the assembled training data, it triggers the problem of similarities or replicated copies of the copyrighted works within the dataset. It is unavoidable to admit that the Large Language Model is built upon a vast dataset comprising millions of images, videos, audio, or text-based works where some of them are reproduced without permission. Consequently, the issue quickly arose regarding whether machine learning should be classified as fair use or subject to other copyright law exceptions. These circumstances could be a legal challenge for startups as the ownership and licensing of datasets are intricate, involving copyrights, intellectual property rights, and privacy regulations. (Lucchi, 2023). Moreover, the expense associated with constructing or obtaining licenses for datasets can be burdensome. Since reproducing copyrighted works could infringe upon the owner’s right to reproduction, this act remains a subject of ongoing legal debate worldwide (Kahveci, 2023). Despite that, copyright, similar to any property right, is not always definitive as it usually incorporates some user privileges or exemptions, such as situations in which the law will permit or authorize specific uses of a copyrighted work even in the absence of permission from the copyright holder. These exemptions depend on various circumstances but often revolve around socially beneficial uses, including education, news reports, academic research, personal development, or public discourse (Burk et al., 2019). Despite the cases above involving different legal actions, they are all claimed under copyright law.
Is training data fair use?
In fact, the above discussion concludes with an important question from both legal and business standpoints: How can we define the training data for fair use in generative AI? Mark Lemley, a prominent expert in intellectual property rights in the United States, who is defending Stability AI (Jackson, 2023). He advocates for the inclusion of copyrighted works in training data as considered fair use in his recent publication “Fair Learning”. Lemley expresses three supportive arguments proving the general use of databases in the AI landscape: 1) The creation of new databases can help enhance AI transparency, 2) It is acknowledged that machine learning is a transformative process where it changes the purposes of the works used, and 3) Required licensing materials for training data is impractical due to the extensive use of such sets (Hayes, 2023). In 2022, machine learning will become available for consumer use, where humans take advantage of creativity to create their own prompts for these computational tools and enable them to learn and reproduce new works from experienced training datasets. The purpose of art and culture is to prosper our social environment and support further innovation. Under this circumstance, the training database is depicted as a consumable product and should be considered fair use because the purpose is non-profit and research oriented.
Alternative solutions for regulating AI training data
To resolve such concerns, the Copyright Office provides an essential strategy to protect the authorship of artworks on the Internet: establishing a contractual relationship. The contractual relationship is conducted between the owners and the users of the works by signing a contract that the Copyright Office legally approves. One example of such a legal contract is the Creative Common (CC) license, where authors allow users to download or copy their works for specific uses if they commit to these licenses. Nowadays, the existing legal framework has offered some direction about using copyrighted material for AI training, such as the TDM exemption in the USA and the EU (Lucchi, 2023). Nevertheless, these frameworks may still need to fully grasp the intricate dynamic natures in the AIGC landscape, where the obstacle of copyright infringement may still occur. The goal of achieving a balance between technological development and the protection of creators’ rights is crucial as the age of AI slowly immerses humans in daily life. Specifically, the public has gained greater attention to transparency and equity in data utilization, and alternative solutions to regulate AI training data have become more important. Strategies such as establishing legal data-sharing contracts or licenses and formulating ethical guidelines or industry standards can help supervise the legal system in AI.
Conclusion
Advanced technologies signify meaningful progress while also implying dangers. The legal system designed to protect human creativity and cultural property has encountered ethical and social problems as Artificial Intelligence develops to a mature stage. Incorporating copyrighted materials into training systems is becoming more prevalent, which urges establishing clearer borderlines to protect intellectual property rights. Looking ahead, concerns such as “what forms of creation should be encouraged”, “which works should be protected”, and “how to regulate freedom of human creativity” are still the main subjects that need to be explored and resolved thoroughly. In consequence, such technological advancements will require a fundamental transformation in our understanding of creativity and a corresponding reassessment of our approach to copyright.
References:
Bao Daming. (2017). The Future of Artificial Intelligence in China. In the China Development Forum. https://www.mckinsey.com.cn/wp-content/uploads/2017/03/CDF_McKinsey_AI_CN_final.pdf
Brittain, B. (2023). Getty Images lawsuit says Stability AI misused photos to train AI. Reuters. https://www.reuters.com/legal/getty-images-lawsuit-says-stability-ai-misused-photos-train-ai-2023-02-06/
Burk, D. L. (2019). Algorithmic fair use. University of Chicago Law Review, 86(2), 283–308.
Flew, T. (2021). Regulating platforms. John Wiley & Sons.
Hayes, C. (2023). Generative artificial intelligence and copyright: both sides of the black box. Social Science Research Network. https://doi.org/10.2139/ssrn.4517799
Jackson, B. (2023, October 5). Consider the risks of generative AI before adopting Game-Changing tools. Forbes. https://www.forbes.com/sites/forbestechcouncil/2023/03/03/consider-the-risks-of-generative-ai-before-adopting-game-changing-tools/?sh=4bef88434aff
Kahveci, Z. Ü. (2023). Attribution problem of generative AI: a view from US copyright law. Journal of Intellectual Property Law & Practice, 18(11), 796–807. https://doi.org/10.1093/jiplp/jpad076
Lucchi, N. (2023). ChatGPT: A case study on copyright Challenges for Generative Artificial Intelligence Systems. European Journal of Risk Regulation (Print), pp. 1–23. https://doi.org/10.1017/err.2023.59
Nelson, M. (2024, January 9). Midjourney AI art program faces lawsuit over alleged use of Magic: The Gathering art. ReadWrite. https://readwrite.com/midjourney-ai-art-program-faces-lawsuit-over-alleged-use-of-magic-the-gathering-art/
The Aesthetic Ethics of Midjourney under the Development of Artificial Intelligence. (2023). Journal of Artificial Intelligence Practice (Print), 6(5). https://doi.org/10.23977/jaip.2023.060507
The problem with plagiarism | Design Education Forum of SA. (2011). https://www.defsa.org.za/papers/problem-plagiarism
“This is gonna change everything”: A very short history of generative AI. (n.d.). Huawei. https://www.huawei.com/en/media-center/transform/12/03-jerry-kaplan Vincent, J. (2023, January 16). AI art tools Stable Diffusion and Midjourney targeted with copyright lawsuit. The Verge. https://www.theverge.com/2023/1/16/23557098/generative-ai-art-copyright-legal-lawsuit-stable-diffusion-midjourney-deviantart
Be the first to comment