Artificial Intelligence and Copyright – Part 2: Absence of Protection for AI-generated Works, Analysis and Reasons of a “Political” Choice

head, wireframe, face-663997.jpg

1. Denial of protection for AI-generated works: the case of “Zarya of the Dawn”

The United States Copyright Office, with its decision on February 21, 2023, denied protection to artificial intelligence (AI) generated images under US copyright law. The case involved Kristina Kashtanova, the author of the graphic novel “Zarya of the Dawn”, who used the software “AI Midjourney” to generate images for her comic.

Initially, the US office had granted registration for the work but then, after learning about the author’s use of the “Midjourney” software, reversed its decision and initiated the process of canceling the registration.

The office contested the way the software generated the images, stating that, in its view, they did not show a substantial contribution from the author, and therefore, she couldn’t claim ownership of those images.

Midjourney is a software that uses text prompts to generate four images in response to user instructions. The user can choose among the images provided by the software or request new variations.

Example of a “prompt” and the result provided by the AI software Midjourney

The office, analyzing the generation process of the images, emphasized that Midjourney does not interpret the prompts as specific instructions to create a particular expressive result because it “does not understand grammar, sentence structure, or words like humans do.”

According to the office, the commands sent by the user function more as suggestions than as actual instructions, generating unpredictable results over which the user has no influence.

In this regard, the US office, referring to the requirement of human creation, highlighted the need to examine the level of human intervention in the processing of the work in order to evaluate its registrability on a case-by-case basis.

In the specific case, the office considered the images to be a mere product of the AI software’s activity, and the author’s subsequent requests for modifications and edits to the images were deemed insufficient to constitute an original creation eligible for copyright protection. These modifications were considered minor and lacking the necessary creativity.

At the same time, the decision of the Copyright Office recognized the authorship of the text of the graphic novel and the arrangement of the images by Kristina Kashtanova, as the office did not find evidence of the use of any generative AI for these activities.

This last aspect of the decision confirms that, regarding composite works, the lack of protection for a compositional element under copyright law does not prevent the protection of the overall work, including its structure and layout.

2. Analysis of a “political” choice

The analyzed case gives rise to several reflections on the relationship between artificial intelligence and copyright law.

i. The issue of “predictability of the result”

If, according to the US office, the unpredictability of the result obtained through the use of expressive tools undermines the protection of the work generated in this way, one wonders how this ruling can cope with the creative process of numerous works of art.

In the art world, to achieve something original and creative, artists often experiment with different means, tools, and methods of creation, which by nature cannot be “predictable.”

An example could be the artist Jackson Pollock, a well-known representative of the “action painting” movement, and his works consisting of paint splatters on canvas.

In these works, the exact rotation radius and trajectory of the paint drops on the canvas were certainly neither known nor predictable to the artist. Furthermore, Pollock’s intention was to capture an image on canvas, but he did not know in advance what the actual result of his actions would be, similarly to Kristina Kashtanova and her desire to generate images for the graphic novel.

The question arising from all this is: “If we follow the reasoning of the US office in this particular case related to the predictability of the expressive means used, how many works of art could be denied copyright protection?”

Convergence, 1952 by Jackson Pollock

ii. The analysis of the level of creativity

The second issue that emerges from the decision concerns the use of different standards for assessing the creative contribution required for copyright protection, depending on the medium of creation used.

In this regard, it is worth recalling the example of photographic works, for which it is common practice to grant protection without a systematic verification of the author’s creative control over lighting, angles, and all other elements that make up the originality of a photograph.

On the contrary, in this case, the US office demonstrated meticulous attention regarding the author’s creative control over the generative process of the images, considering neither the author’s initial instructions for image generation, nor the selection process carried out by the author, nor the subsequent requests for modifications as significant enough for protection.

In relation to this aspect, instead of questioning the amount of creative contribution necessary for copyright protection, it would be more appropriate to focus on the reasons behind the “rigidity” demonstrated by the offices.

Indeed, the level of interaction between the user and the AI software required for copyright registration appears to be particularly high. This leads to the assumption of a discouraging intent from the US office towards the use of Artificial Intelligence as a tool for creating intellectual works.

This reluctance is understandable and probably justified by the consequences that granting copyright to AI-generated works could have.

Certainly, the well-known question of the ownership of such rights would arise: the user of the AI (who merely provides the “prompt” commands), the creator of the AI (the one who conceived its functioning), the owner of the AI, and the AI itself (if a legal framework governing rights and obligations were established).

Furthermore, and above all, the granting of such rights could risk saturating the art market. In this sense, as software can generate a multitude of images quickly, the automatic protection of these images would flood the market with rights, making it extremely difficult for culture to spread and for works to freely circulate, betraying the essence of copyright.

As always, the choice is ours. However, between protection and the public domain, nothing excludes the possible adoption of a third way that goes beyond the traditional canons of copyright law.

Leave a Comment

Your email address will not be published. Required fields are marked *