Image from here. Théâtre D’opéra Spatial.
On September 5, 2023, as explained here, the US Copyright Office (USCO) issued an interesting decision in a copyright registration matter that involved AI-generated work. Previously, in the Thaler case, the US Copyright Office had refused to register an AI-generated work since the application named the AI-system as the author. This case revolved around the question of whether or not an AI system can be an author. The USCO ruled that it cannot since human authorship is necessary.
Unlike the Thaler case, artist Jason Allen had applied for copyright registration of an AI-generated work (Théâtre D’opéra Spatial) naming himself as the author. The work was generated by the Midjourney image AI system. The work has garnered national attention for being the first AI-generated image to win the 2022 Colorado State Fair’s annual fine art competition.
A few days back, the USCO refused to register this work as well. The grounds for refusal were different from the Thaler decision because authorship was not the issue here. The issue the USCO found with this application was that it failed to meet the de minis standard and the author had failed to disclose and disclaim the AI-generated portions of the work. They found that the AI-generated amount of content was more than de minimis and the applicant sought to register the entire work refusing to disclaim the portions attributable to AI.
The applicant had filed his copyright application without naming Midjourney as the author and also without disclosing the fact that the work was AI-generated. However, since the work had won a national award, the copyright office was aware that it was AI-generated.
Apart from this, the applicant argued that he contributed significantly to the creation to this image. His ‘creative inputs’ included providing around 624 text prompts and revisions of text prompts and using software to remove flaws and upscale the image. In relation to the prompts, the applicant argued that he dictated the tone of the image:
“As explained in his correspondence, Mr. Allen created a text prompt that began with a “big picture description” that “focuse[d] on the overall subject of the piece.” [..] He then added a second “big picture description” to the prompt text “as a way of instructing the software that Mr. Allen is combining two ideas.” Id. Next, he added “the overall image’s genre and category,” “certain professional artistic terms which direct the tone of the piece,” “how lifelike [Mr. Allen] wanted the piece to appear,” a description of “how colors [should be] used,” a description “to further define the composition,” “terms about what style/era the artwork should depict,” and “a writing technique that Mr. Allen has established from extensive testing” that would make the image “pop.” Id. He then “append[ed the prompt] with various parameters which further instruct[ed] the software how to develop the image,” resulting in a final text prompt that was “executed . . . into Midjourney to complete the process” and resulted in the creation of the Midjourney Image above.”
However, the USCO was not convinced. The USCO cited its AI Registration Guidelines (Copyright Registration Guidance: Works Containing Material Generated by Artificial Intelligence, 88 Fed. Reg. 16,190, 16,192 (Mar. 16, 2023)) to assess whether a work’s “traditional elements of authorship” were produced by a machine or a human. In this case, though the applicant provided hundreds of prompts before the image was finally created, the steps in the process were ultimately dependent on how the Midjourney system processed Mr. Allen’s prompts. No matter how creative the inputs, the output was a result of interpretation by Midjourney based on its training data set. The USCO ruled that prompts are not ‘direct instructions’ which Midjourney understands like humans do. On this basis, the USCO ruled that more than de minis of the work was determined and executed by technology and not by a human.
Prompts, AI, computer generated works and Indian law
While the issue around copyrightability of AI-generated works is far from clear in the Indian context, it is interesting to examine the impact of this line of reasoning under Indian copyright law. (See my post here on AI and copyright registration issues in India).
The Copyright Act, 1957 defines an author in relation to any literary, dramatic, musical or artistic work which is computer-generated as “the person who causes the work to be created”. The term ‘computer generated work’ is not defined and is to be understood literally. Regarding the requirement for human authorship, it has been argued, that the Copyright Act requires a human author for computer-generated works because the term of copyright for such works is calculated using the age of a natural person. There is no Indian precedent to suggest that authorship can be conferred to computer / software. There is also no policy guidance from the Indian Copyright Office on whether only humans can be considered authors.
Further, neither the India Copyright Office nor Indian courts have dealt with how the words ‘the person who causes the work to be created’ should be interpreted. Is it the person who creates the AI-system that causes the work or is it the person who provides the prompts (without prompts there is no work)? But based on the USCO line of reasoning, it could be argued that human inputs/ prompts, however creative, are not direct instructions that ‘cause’ the ultimate work to be created. Such prompts only ‘influence’ the AI-system and the ultimate output is a result of the AI-system’s own interpretation of the prompts based on its training data set.
AI-generated images have several applications in various fields from digital art, simulations for training purposes, marketing campaigns, personalized avatars, 3D models etc. Given its myriad applications, the way courts decide copyright issues over such content will impact people hoping to protect AI-generated content.