The Misguided Stance of the Copyright Office on AI-Generated Works

Many practitioners assumed the issue of authorship would be the primary focus of the copyright office’s initiative to examine “copyright law and policy issues raised by artificial intelligence (AI).”  Unfortunately, this is not the case. Instead, Part 1 of the copyright office’s Report on this initiative Digital Replicas Report addresses deep fakes. 

Yes, deep fakes are a problem. However, this is not a copyright problem, and the report offers no new insights. Shira Perlmutter, Register of Copyrights and Director of the U.S. Copyright Office, accurately sums up the report in two sentences:

 

“It has become clear that the distribution of unauthorized digital replicas poses a serious threat not only in the entertainment and political arenas but also for private citizens. We believe there is an urgent need for effective nationwide protection against the harms that can be caused to reputations and livelihoods.”

 

How insightful…

In this post, point out obvious problems with the Copyright Office’s position surrounding authorship of works generated using AI tools and why there is no hope of this position being applied consistently across all digital images. 

Copyright Law: The Basics

Copyright is simple: The creator has ownership of the creation.

There are only questions relevant to the process of registering a copyright are: 

 

1. Did the creation exist before the creator created it?  

2. Did the creator create it (in copyright terms, is the creator the “author”)?  

In the vast majority of cases, the answers to these questions are self-evident and the copyright is granted.  Simple.

AI and the Authorship Dilemma

The use of AI image generating tools, like Midjourney, in the creative process has sparked concern. The U.S. Copyright Office has taken the position that the tool (not the user using the tool) is the author. The best example we have is the case of  Kristina Kashtanova’s graphic novel, Zarya of the Dawn, which was partially generated using AI. The U.S. The Copyright Office initially granted her copyright protection but later revoked copyright protection for the AI-generated images in the work for lacking human authorship. 

This reasoning is flawed. Since authorship is a core focus of copyright law, the issue of who the author is when using AI tools needs careful examination.

 

1. Burrow-Giles Lithographic Co. v. Sarony

In both its policy guidance regarding AI and its letter to Kashtanova denying the copyright after appeal, the Copyright Office cites Burrow-Giles Lithographic Co. v. Sarony (1884) as the primary source discussing the meaning of “authorship.” 

Burrow-Giles involved a photographer, Napoleon Sarony, who sought to register his photographs of the famous actor Oscar Wilde. The defendant argued that photographs were mere mechanical reproductions and thus should not qualify for copyright protection. The Court ruled in favor of Sarony, holding that photographs are indeed the product of human intellectual effort and creativity, “so far as they are representatives of original intellectual conceptions of the author.”  

Oscar Wilde by Sarony

 

2. The Copyright Office’s “Authorship Rule”

The Copyright Office’s “authorship rule” defined by their AI policy guidance echos Burrow-Giles, asking “whether the ‘work’ is basically one of human authorship, with the computer [or other device] merely being an assisting instrument, or whether the traditional elements of authorship in the work (literary, artistic, or musical expression or elements of selection, arrangement, etc.) were actually conceived and executed not by man but by a machine.” 

Where Does Predictability Come Into The Picture…

In the Kashtanova case, the Copyright Office concluded “it was Midjourney—not Kashtanova—that originated the “traditional elements of authorship” in the images.” This infers that the images were MidJourney’s “original intellectual conceptions,” which is definitely not the case

Midjourney did one thing: It translated Kashtanova’s description of an image into an actual  image. That’s it.  

Midjourney could never have produced the images used in Kashtanova’s graphic novel de novo. Therefore, Midjourney is not capable of “original intellectual conceptions.”  Because can’t conceive of an image, Midjourney is NOT capable of being an author under the rule defined by the court in Burrow-Giles or the Copyright Office’s authorship rule. 

The Copyright Office attempted to overcome this deficiency by focusing on “predictability,” which is not a “traditional element of authorship” like “conception” and “execution.”  In fact, predictability is not even mentioned in the Copyright Office’s AI policy guidance.  

Victorian Cameras

Focusing on “predictably” in this case was a mistake as Victorian era cameras were not known for their predictability. 

When Burrow-Giles was argued in 1884, cameras primarily used photographic plates coated with light-sensitive chemicals. These early cameras were large, often mounted on tripods, and required long exposure times due to the slow sensitivity of the plates. After a photograph was taken, the glass plate had to be developed in a darkroom, where it was exposed to a series of chemical baths to reveal the image. These steps included developing the latent image with a developer solution, stopping the process with a stop bath, and then fixing the image with a fixer solution. Once fixed, the plate was washed and dried, producing a negative image that could be contact-printed onto photographic paper to create positive prints.

Sarony circa 1899

Getting a good picture required: (1) Leaving the hand controlled aperture open for the appropriate amount of time (the emulsion on photographic plates was not standard, meaning the right exposure time varied from plate to plate), (2) Photographing a subject that doesn’t move while the aperture is open, (3) Having the appropriate chemicals in the appropriate concentrations for development, (4) Timing the transition between developer solution and stop bath exactly correct, and (5) Exposing photographic paper for the appropriate amount of time. Software handles most of these issues now, but in 1884, the photographic process was extremely unpredictable.

Comparing Victorian Photography to Using Midjourney 

Sarony was obviously a master of his craft (and extremely lucky). Kashtanova is no Garth Ennis, but her creative process mirrors Sarony’s. Kashtanova composed the image in prose, directing Midjourney what elements to include in the image, where the elements were to be positioned, and the style and tone of the image. Sarony, similarly, positioned Oscar Wilde, placed the various other elements of image around him, and undoubtedly instructed him to “stay still.” This is “conception.”   

Representative Image from Zarya of the Dawn

Kashtanova hit enter. Sarony physically opened the aperture of the camera. Both Kashtanova and Sarony hoped that the image they imagined, i.e. their “original intellectual conceptions,” matched the actual image produced when processing was complete. This is “execution.” 

The “traditional elements of authorship” lead to Kashtanova being the author. The only difference between Kashtanova’s images and Sarony’s photograph is how they were processed. Sarony used chemistry. Kashtanova used a computer. 

An Ambiguous Rule That Will Lead To Uncertainty

What metrics should practitioners use to judge the “predictability” in these cases? The Copyright Office’s AI policy guidance doesn’t answer this question. The reasoning in the Kashtanova case is so convoluted that it is not helpful, and Part 1 of the Copyright Office’s AI Initiative Report is focused on something entirely different. 

Millions of copyrighted images have been processed using computers. Authors of copyrighted images filter and manipulate those images using AI. The copyright office does not question authorship in these cases. Nor, does the copyright office question the authorship of CGI sequences in movies either even though AI is used extensively in motion capture, facial animation, and crowd simulations among other things. When is the “predictability” rule applied and when is it not?

Conclusion

The court in Burrow-Giles understood that artists like Sarony and Kashtanova would embrace technology and incorporate it into their creative processes. Embracing technology, the rule they defined incorporated an understanding of the artistic process: conceiving of a work and executing that vision. Kashtanova meets this standard. The Copyright Office used short sighted reasoning and distorted facts to arrive at an unwarranted and factually inaccurate conclusion. The decision in Kashtanova cannot stand.

Next
Next

Beware of Trademark Scams: How to Spot and Avoid Deceptive Tactics Targeting Your Business