Can you copyright generative AI? The answer surprisingly originates with the camera, in a key US Copyright Office report

AI lawsuit over copyright infringement
(Image credit: Beth Nicholls)
Key Points from the US Copyright Office report

• Works created entirely by prompts and generative AI cannot be copyright protected

• Works created by both humans and AI need to be determined on a case-by-case basis

• The copyright of photographs was once questioned because they "were the product of a machine.”

Existing intellectual property law is enough to guide artists into the age of generative AI, the US Copyright Office has concluded in a new report published on January 29. In the second part of the Artificial Intelligence Report, the US Copyright Office addresses key questions over whether or not generative AI can be protected under current copyright law. The report says that prompt-based generative AI cannot be copyrighted, but addresses the questions surrounding works made by a joint effort between humans and AI.

The report is key for artists because it outlines what kind of generative AI can and cannot be copyrighted. Writing a prompt isn’t enough human involvement to file for a copyright, the report says. Works created with a mix of human and AI involvement will be determined on a case-by-case basis, although the report offers some insight into how much AI is too much AI. But the legal precedence for the Copyright Office’s conclusions actually date back more than 100 years to the invention of the camera, when the Supreme Court considered whether a photograph was copyrightable or “the product of a machine”.

One of the key questions the report addresses is whether or not existing law works with the new and growing technology. The report says existing law is enough, but one of the legal precedents that helped the US Copyright Office come to that conclusion dates back to 1884.

Daguerreotype photography was invented by the French photographic pioneer Louis Daguerre (1789-1851) and was made public in 1839. In the Daguerreotype process a picture made on a silver surface sensitized with iodine was developed by exposure to mercury vapour. Daguerre granted the right to make and sell daguerreotype cameras to a relative of his, Alphonse Giroux of Paris. Photography thus became available to the general public in 1839 and this was the first commercially available camera, which produced the first distinctive photographic positives. It took pictures 8.5 x 6.5 inches (21.6cm x 16.5cm), a size which became known as whole-plate when later cameras were built to take photographs a fraction of the size. (Photo by SSPL/Getty Images)

(Image credit: Science & Society Picture Library / Getty Images)

“More than a century ago,” the report reads, “the Court analyzed the nature of authorship in a case involving the then-new technology of the camera. In Burrow-Giles Lithographic Co v Sarony, the Court considered a constitutional challenge to Congress’ extension of copyright protection to photographs. The defendant argued that photographs were not copyrightable because they lacked human authorship; instead, they were the product of a machine.”

In that case, the courts decided that the posing, arranging the subjects and prompting an expression was enough of a human element to grant copyright protection to photographers. That human element is key, as the report also mentions the 2014 US Copyright Office decision that a selfie taken by a monkey isn’t eligible for copyright protection.

The legal precedence from the invention of the camera, along with existing legislation, is enough to address the question of a new form of works created by a machine: generative AI. The camera is a machine, but the photographer has enough creative control over what is in the frame to warrant copyright protection. Similarly, generative AI may only qualify for copyright protection if there is human authorship. But how much of a human element is needed? The report then outlines the debate and answers key questions on whether or not generative AI can be copyrighted.

AI (Artificial Intelligence) concept - woman creating photos from her thoughts

(Image credit: Getty Images)

While the US Copyright office says that works created with a mix of AI and human authorship will need to be determined on a case-by-case basis, the answer on works generated entirely with a written prompt is more clear cut.

“The Office concludes that, given current generally available technology, prompts alone do not provide sufficient human control to make users of an AI system the authors of the output,” the report says. “Prompts essentially function as instructions that convey unprotectible ideas. While highly detailed prompts could contain the user’s desired expressive elements, at present they do not control how the AI system processes them in generating the output.”

The report then looks at the gaps between the prompt and the result, using an example of a cat reading a newspaper generated with Google Gemini. The resulting image missed some of the prompt instructions, and added others, like human hands holding the newspaper.

The fact that the same prompt given multiple times will never generate the same exact image further illustrates the lack of human control over generative AI, the report suggests. Even rewriting the prompt multiple times isn’t enough to create sufficient human authorship, the report says.

The report then looks at some of the public input on the topic and, in particular, the correlation between the seemingly random results of generative AI and some types of artwork that also include an element of randomness, such as a Jackson Pollock style painting where the paint is distributed at random or a wildlife camera where what stumbles in front of the lens is just random chance. However, the US Copyright Office says that those types of work still have a human author behind the expressive elements.

“Jackson Pollock’s process of creation did not end with his vision of a work,” the report says. “He controlled the choice of colors, number of layers, depth of texture, placement of each addition to the overall composition – and used his own body movements to execute each of these choices. In the case of a nature photograph, any copyright protection is based primarily on the angle, location, speed, and exposure chosen by the photographer in setting up the camera, and possibly post-production editing of the footage.”

Can works made in part by AI be copyrighted? It depends

The Photoshop software and logo is displayed on a computer monitor

(Image credit: Getty Images / Mahmut Serdar Alakus / Anadolu Agency)

While the report made it clear that prompt-based generative AI is not eligible for copyright protection, works that are created in part by a human artist and in part by generative AI are less clear cut. “Copyright does not extend to purely AI-generated material, or material where there is insufficient human control over the expressive elements,” the report says. “Whether human contributions to AI-generated outputs are sufficient to constitute authorship must be analyzed on a case-by-case basis.”

While the answer for partially AI-generated works will be decided on that basis, some of the examples offered by the Copyright Office offer some insight into what may qualify for legal protection. The Office granted copyright to a work that was created when the artist uploaded a hand-drawn illustration to an AI system and asked for specific modifications. The resulting work had the original drawing’s composition, lines and arrangement, but the AI added a photorealistic style. However, that copyright was granted with an annotation that the copyright is “limited to unaltered human pictorial authorship” and doesn’t extend to what the AI generated.

The Copyright Office also accepted the copyright application for a comic book that was written by a human author but illustrated by AI, determining that the text and arrangement on the page was enough to grant copyright.

The report then discusses tools that allow humans greater control over generative AI, such as Midjourney’s Remix Prompting, but says, again, whether or not the work has enough original creative human element would need to be determined on a case-by-case basis.

The examples offered seem to suggest that an original photograph that is edited by AI could be protected under copyright law at least in part, but again, the US Copyright Office says that mixed works will be determined on a case-by-case basis. “[T]he inclusion of elements of AI-generated content in a larger human-authored work does not affect the copyrightability of the larger human-authored work as a whole. For example, a film that includes AI-generated special effects or background artwork is copyrightable, even if the AI effects and artwork separately are not.”

How will the report affect AI companies? Hugging Face, an AI company, seemed to suggest that the lack of copyright protection won’t drastically impact the technology. “[W]e find that very little to no innovation in generative AI is driven by the hope of obtaining copyright protection for model outputs,” the AI company is quoted as saying in the report. “The incentives for innovation already exist without modifying copyright law.”

You may also like

For more, read our guide to copyright for photographers, or research the ongoing battle on using copyrighted images to train AI.

Hillary K. Grigonis
US Editor

With more than a decade of experience reviewing and writing about cameras and technology, Hillary K. Grigonis leads the US coverage for Digital Camera World. Her work has appeared in Business Insider, Digital Trends, Pocket-lint, Rangefinder, The Phoblographer and more.

You must confirm your public display name before commenting

Please logout and then login again, you will then be prompted to enter your display name.