By Eloise Wilkes-Barnwell

The rapid evolution of generative AI technologies has transformed the landscape of creativity, enabling machines to produce text, images, music, and more with unprecedented complexity. Nevertheless, this evolution presents substantial challenges for existing copyright laws which were not designed to anticipate such innovations.

As legal systems globally grapple with these complexities, the urgency to clarify the law’s stance becomes increasingly critical to protect the rights of creators and innovators in this digital age. This article examines the intersection of generative AI and copyright infringement, highlighting the pressing legal issues through an exploration of recent copyright infringement cases involving the use of generative AI.

Currently, two primary issues are at the forefront of discussions regarding copyright infringement. First, the inclusion of copyrighted material in generative AI databases for training (input); and second, the copyright status of the content generated by generative AI (output).

Traditionally, copyright law has required human authorship to be eligible to attain copyright protection. A notable precedent on this topic is the “monkey selfie” case (Naruto v Slater), where the question of authorship arose due to a photograph taken by a non-human entity, a monkey. This case reinforced the notion that works created by non-humans cannot be copyrighted.

Navigating copyright principles in the context of generative AI presents increasing complexities regarding how copyright law should address situations where AI systems autonomously generate content based on human inputs. Recent legal cases and varying jurisdictional approaches offer perspectives on these challenging issues. They delve into the legality of using datasets containing copyrighted material without proper authorisation, highlighting potential ramifications for copyright holders. Additionally, these discussions explore the fundamental query of ownership: whether individuals can rightfully claim copyrights for the outputs produced by generative AI.

Cases Decided

• Zarya of the Dawn (US Copyright Office)

In a recent case involving the graphic novel “Zarya of the Dawn,” the US Copyright Office (USCO) were presented with the question of whether images generated with the help of the AI system, Midjourney, could qualify for copyright protection. Initially registered by Ms. Kristina Kashtanova without disclosing the AI’s involvement, the novel prompted the USCO to investigate the nature of the AI-generated content.

The USCO concluded that images created by Midjourney did not meet the criteria for copyright protection. It determined that Midjourney’s process, which generates images based on text prompts without predictable outcomes, lacked the human authorship required under copyright law. Unlike tools under human control, Midjourney’s operations did not involve guided creative decision-making by a human artist. Therefore, the USCO denied copyright registration for the AI-generated images due to their lack of originality. This decision highlights the USCO’s stance that copyright law reserves protection for works originating from human creativity.

• Thaler v Perlmutter (U.S)

In a significant legal matter heard by the U.S. District Court for the District of Columbia, the court addressed a dispute involving the U.S. Copyright Office (USCO) and Stephen Thaler regarding the registration of an AI-generated visual work. Thaler claimed that the work, autonomously created by an AI algorithm named Creativity Machine, should be eligible for copyright protection, despite lacking direct human authorship. The USCO initially rejected Thaler’s copyright application in August 2019, citing the requirement for human involvement in the creative process.

Thaler subsequently filed a lawsuit under the Administrative Procedure Act, challenging the USCO’s decision and arguing that as the owner of the AI system, he should be entitled to copyright ownership of the generated work. However, the USCO maintained its position, asserting that the absence of direct human authorship precluded the work from copyright protection. The central issue in this case revolved around whether an AI-generated work could be considered copyrightable under current legal frameworks.

•  Li v Liu (China)

In a notable copyright case heard by the Beijing Internet Court in China, the court deliberated on the protection of AI-generated images and ownership of their copyrights. The dispute arose when the plaintiff utilised the AI model Stable Diffusion to create an image of a young woman. Subsequently the Defendant posted the image without authorisation on a social media platform, leading to a claim of copyright infringement by the plaintiff.

The court affirmed that AI-generated works could in fact be eligible for copyright protection under Chinese law, provided they meet the criteria of originality and intellectual achievement. In this instance, the plaintiff demonstrated intellectual input by designing the appearance of the woman in the image through detailed prompts and adjustments, reflecting subjective aesthetic choices and original judgment.

Regarding copyright ownership, the court ruled that the AI model itself cannot be considered the author or owner of copyright because it lacks legal personhood under copyright law. Instead, the court recognised the plaintiff, who actively contributed to the creative process and held responsibility for the design and adjustments made using Stable Diffusion, as the rightful author and copyright owner of the image.

This case highlights China’s evolving stance on AI-generated content, indicating a willingness to recognise copyright protection for such works when human creativity and judgment play a substantial role in their creation.

• Ultraman v AI company (China)

The Guangzhou Internet Court in China ruled on a significant copyright infringement case involving outputs generated by AI. The case centred on a defendant operating an AI-powered text-to-image generator that allowed users to create images resembling copyrighted works from the Ultraman series, owned by Tsuburaya Productions Co., Ltd. Shanghai Character License Administrative Co., Ltd. (SCLA), which holds licensing rights for Ultraman in China, sued the defendant for infringing its exclusive rights to reproduce and create derivative works based on Ultraman imagery.

The court ruled that the defendant’s AI-generated images reproduced substantial elements of the copyrighted Ultraman works. The court emphasised that the defendant’s platform enabled users to generate images “identical or substantially similar” to the copyrighted material by using the keyword “Ultraman.”

Additionally, the court also determined that the defendant created unlawful derivative works by producing outputs that incorporated elements of the original Ultraman imagery without authorisation. This case sets a precedent in China’s legal approach to AI-generated content and copyright infringement. It highlights the importance of AI service providers complying with regulatory frameworks and implementing safeguards to protect intellectual property rights.

Cases in Motion

• Getty Images v Stability AI (UK)

The legal dispute between Getty Images and Stability AI, currently before the High Court in London, revolves around allegations of copyright infringement and related intellectual property violations. Stability AI, a London-based developer, is accused by Getty Images of unauthorised use of its images for training AI models. Getty alleges that Stability AI’s actions violate copyright, database rights, trademarks, and laws against passing off. The primary legal issue in question is whether Stability AI’s activities, particularly its use of Getty Images without authorisation to train its AI system and create AI-generated outputs resembling copyrighted works, constitute copyright infringement under UK law.

The case is set for trial in summer 2025 after the High Court rejected Stability AI’s application to strike out certain claims brought by Getty Images. The outcome of this case could significantly influence future copyright licensing regulations in the context of AI technologies in the UK.

• Anderson v Stability AI (U.S)

A California federal judge allowed a significant copyright lawsuit against Stability AI, Midjourney, and other entities accused of using artists’ work without authorisation to train their artificial intelligence-based image generation systems. The lawsuit, initiated by ten artists including Sarah Andersen, Kelly McKernan, and Karla Ortiz, alleges that these companies, which include DeviantArt and Runway AI, unlawfully copied and stored their artworks on company servers for the purpose of training AI models.

The judge’s ruling indicates that the artists have presented plausible arguments for copyright infringement, paving the way for the case to proceed to the next stage of litigation. This case highlights the emerging legal complexities surrounding AI technology and copyright law, raising questions about the permissible uses of copyrighted material in machine learning and artificial intelligence development.

• Authors Guild v OpenAI (U.S)

Multiple copyright holders, including the Professional Writers Association, individual authors, and a news organization, have sued Microsoft Corporation and OpenAI, alleging that OpenAI’s ChatGPT product unlawfully used their copyrighted materials to train its Large Language Models (LLMs), resulting in the generation of infringing text. These lawsuits have been consolidated into four actions before the Southern District of New York, seeking damages for copyright infringement and violations of the Digital Millennium Copyright Act (DMCA). The central issue in the dispute is whether OpenAI’s use of copyrighted works to train ChatGPT qualifies as fair use under U.S. copyright law.

• The Intercept Media, Inc v Open AI (U.S)

Three news organizations—The Intercept, Raw Story, and AlterNet—have filed lawsuits against OpenAI and Microsoft in the Southern District of New York, alleging copyright infringement. The lawsuits claim that OpenAI used thousands of their articles to train its ChatGPT AI system without permission, leading to the reproduction of copyrighted material without proper attribution or authorization. They specifically accuse OpenAI of violating the Digital Millennium Copyright Act (DMCA) by removing essential copyright information from their articles to facilitate this infringement.

Implications

The intersection of generative AI and copyright law presents both significant challenges and opportunities in the digital age. As AI technologies continue to evolve, creating complex works of art, music, and literature, they also test the boundaries of legal frameworks originally designed around human creativity and authorship. These cases highlight the difficulties in addressing issues such as the use of copyrighted materials in AI training datasets and the copyright status of AI-generated outputs. They also underscore the urgent need for clearer copyright laws to protect creators’ rights while fostering innovation in AI development.

Beyond the legal implications, these cases spark broader societal discussions on the ethical and economic impact of AI on creative industries. They challenge us to rethink the very notions of authorship and creativity in an increasingly automated world.