Georgetown Law
Technology Review

The name “text-to-image AI-generator” may not turn heads, but the technology it describes does. Text-to-image AI-generators are artificially intelligent programs that create art. Programs like Stable Diffusion, Artbreeder, and DALL-E scrape the Internet for art, formulate datasets, and draw from those datasets when users prompt the program to “create” artwork. Think ChatGPT, but the programs spit out artwork instead of text blurbs. 

These programs present a concrete problem for artists: they threaten artists’ livelihoods and disrupt the future of art more generally. Boris Eldagsen recently illustrated these threats when he won a Sony World Photography Award and later revealed that his submission was created using DALL-E. 

Generative artificial intelligence (AI) also presents legal conundrums from a copyright perspective. Do the artworks that AIs produce infringe on the intellectual property rights of the artists whose work they scrape? Who has rights to and responsibility over the work that the AIs produce? 

Artists are beginning to use copyright—and tech—as tools to fight back.

The Plight of the Artist

AI generators are forcing artists to shift their thinking concerning consent. Disney illustrator Hollie Mengert, was disappointed to learn that her work was used in an experiment to train a machine learning model. Although she spent years refining her craft, the AI generator produced new art using her work in mere hours and without her approval. Training of, and art production by, AI generators without the consent of the original artists make it harder for artists to refine their brand and can even create consumer confusion as to what is or is not their work. 

The surest way for artists like Mengert to evade exposure to generative AI is to refrain from posting their artwork online, which would disrupt traditional channels for advertising their work. But AI is disrupting artists’ workflow even if they do advertise on the web. For example, French publications such as Regards and So Foot are increasingly turning to AI generated images to reduce the cost of image procurement when they cannot find the right artwork to pair alongside their literary work. This changing landscape can leave artists scrambling to adapt.

Protection for Artists

U.S. copyright law does not protect an idea; it protects the expression of that idea. Often a new expression of an idea—in this case, new art—is derived from preexisting works and is therefore considered to be a “derivative work.” A derivative work is entitled to independent copyright protection if the derived product represents an original work of authorship.

Artists suing Stability.AI, the company behind Stable Diffusion, argue that it has unjustly profited from artists’ works without compensating those who hold the copyright of the programs’ training data. According to analysts, one unlikely but potential outcome is that the suit forces tech platforms to treat training data files like digital contraband, bringing an end to generative AI in a decentralized form. If this argument is successful, only centralized generative AI operated by closed companies and models with locked APIs, such as OpenAI, would remain legal.

Artists are turning to solutions outside of the legal landscape, too. Glaze is an application that enables artists to apply “style cloaks” to their art before sharing it online, ensuring that AIs do not pick the art up when they crawl the web.

Another tool that can help artists in copyright claims is HaveIBeenTrained, a website that enables artists to search for their work in the dataset that Stable Diffusion used. This information arms them with the status of their work and their potential for involvement in a class action suit. 

Ultimately, the optimal protection for art and artists dealing in cyberspace mixes copyright and technological innovation. A combination of these tools provides our best chance at ensuring the soundest protection for artists’ intellectual property in the face of A.I.

The Battle Will Rage On

Art created by AI represents the latest in the constant tug-of-war between copyright and new technology. Both are weaponized against each other to protect innovation.

In this context, it is easy to side with artists against intellectual-property-stealing machines. But even this seemingly black-and-white example is nuanced. AI-generators enable researchers to analyze trends in art. They enable the creation of new art by machines for human consumption. They can also inspire artists to move in a new direction while remaining within their style or enable new artists, who otherwise may have been priced out of the art world, to enter the scene.

Art is not static, and while the starving artist is a sympathetic actor next to the machine, it is important to remain open-minded about the importance of the push-and-pull of copyright and technology in creation.