All the news on AI seems to focus on new features and the eventual release of OpenAI’s GPT-5, which promises (through the rumor mill) a significant upgrade in capabilities. But lurking in the background of AI developments are regulatory initiatives and multiple copyright cases filed by artists, authors, newspaper publishers, programmers, and others. Whether these legal moves can derail the AI juggernaut is an open question.

We haven’t been optimistic about the success of these legal actions as the courts have generally upheld the scraping of publically accessible websites. You need only look at the 2022 decision of the U.S. Ninth Circuit of Appeals, which upheld web scraping in a case involving LinkedIn.

. . . the U.S. Ninth Circuit of Appeals ruled that “scraping” publicly accessible information from websites does not violate the Computer Fraud and Abuse Act (CFAA). The case, Hiq Labs, Inc. v. LinkedIn Corporation, grew out of LinkedIn’s efforts to prohibit Hiq from capturing information found in users’ public profiles, and has been closely watched by Web archivists and academics who incorporate Web data into their research projects.

But the latest court ruling in an AI copyright case is being celebrated by artists. In a decision last week, a California judge allowed ten visual artists to move forward with a copyright case against four major companies employing text-to-image generative AI. The artists, who filed a class-action lawsuit last November, allege that their works have been exploited by Stability AI, Midjourney, DeviantArt, and Runway for training their image-generating AI, Stable Diffusion, without consent.

This could be a pivotal moment in the legal battle artists are fighting to protect their work. If they win in the end, it could profoundly impact AI developments.

This Copyright Case Could Reshape the Landscape of Art and AI

Let’s unpack this further, as it’s critical for the future of AI and creative labor.

The crux of the artists’ lawsuit is that Stable Diffusion uses their artwork from a training dataset (more on that in a minute) and replicates their distinct artistic styles, potentially undermining their livelihoods. District Judge William Orrick’s recent decision dismissed claims of unjust enrichment and breach of contract but upheld the plausibility of the artists’ copyright case. This legal acknowledgment sets the stage for a trial that could have far-reaching implications for artists everywhere.

As the case progresses, it brings to the forefront the ethical and legal challenges surrounding the use of AI in creative fields. The basic argument at stake here, according to Philippa Loengard, Executive Director of the Kernochan Center for Law, Media, and the Arts at Columbia University, is that:

You cannot copy someone’s work without their permission and you can’t then keep that copy because that is infringement.

This simple yet powerful statement underscores the fundamental rights of artists at stake in the age of AI. Perhaps most galling for creatives is seeing derivatives of their work in the output of these AI-based platforms. For instance, New York illustrator and vocal AI critic Molly Crabapple has long warned about the destabilizing effects of AI on working artists. In an open letter to the publishing industry, Crabapple urged a boycott of AI-generated illustrations, highlighting how these technologies, built on what she describes as “stolen work,” threaten to erode artists’ livelihoods for the financial gain of Silicon Valley corporations.

The case centers around the alleged use of the LAION dataset—a massive collection of five billion images—to train Stable Diffusion. This dataset reportedly includes works from prominent artists like David Hockney and Yayoi Kusama, further intensifying the debate over AI’s role in creative industries.

The LAION Dataset At The Core Of The Copyright Case

Numerous issues surround the LAION (Large-scale AI Open Network) dataset from a German non-profit, which is at the core of the copyright case here. The dataset, first released in August 2021, mimicked a model initially created by OpenAI. The current LAION dataset LAION-5B, released in March 2022, contains over 5 billion image-caption pairs. It’s important to note that the dataset doesn’t actually hold the images, but URLs to the images on the web. The captions for the images were generated automatically by the associated ALT tags on web pages. LAION-5B remains a significant resource in the field of artificial intelligence, especially since it is one of the few open-source large-scale datasets freely available, not locked behind corporate walls. It’s widely used for training and testing large-scale machine learning models, especially in natural language processing (NLP) and computer vision.

So if your artistic work is in LAION-5B—and there’s an excellent chance it is if you posted it on the web—what does that mean for your future as an artist? In many ways, the artists’ copyright case revolves around what constitutes a derivative work and at what point the original artist is due compensation. Trained on LAION-5B, an AI image-generating platform doesn’t reproduce the artist’s work nor point to the original source, even though the URLs could be made available. But it does let users generate new work based on this massive dataset of images and captions.

The Possibility of Stolen Livelihoods

From the viewpoint of many artists, the rapid evolution of AI models and their ability to replicate artistic styles poses an existential threat to their livelihood and the creative economy. The potential harm is not limited to high-profile artists but extends to countless illustrators and visual creators who rely on their unique styles for income.

Molly Crabapple goes on to argue,

These aren’t people who made tons of money . . . These are people in their later middle age in America who are facing retirement with nothing because these Silicon Valley companies decided they needed to make billions and billions of dollars.

While this artist copyright case represents a significant step in addressing the legal complexities of AI-generated content, the broader question remains: Can the existing legal framework protect artists in a world where AI increasingly blurs the lines between inspiration and infringement? Could a structure of compensation be built in? Given that the images have already been scraped and used in AI-based creative works, it is hard to imagine how this would be done.

How this copyright case plays out is anyone’s guess, but it will set a precedent that either safeguards the rights of creators or further erodes them. It could also have a major impact on the future development of AI image-generating platforms.

Stay tuned, as this copyright case is only the beginning of the Pandora’s Box of ethical quandaries created by the explosive development of generative AI.