The pipe-smoking cat judging the place of AI in copyright
A studious new mascot for the U.S. Copyright Office
The U.S. Copyright Office is currently in the process of releasing its three-part report on artificial intelligence’s influence on copyright. While not law, it provides authoritative guidance on how AI usage can impact whether an image, composition, film or music can be attributed to a human author.
Digital Replicas was published in July 2024, with a focus on AI-generated copies and deepfakes, providing seminal guidance that will impact news and entertainment industries like film and music.
Copyrightability was published last week, and detailed whether or not a work created with AI can by attributed to the artist. The report is stirring conversation because it states that a work created with AI can by copywritten, so long as their creative impression is evident, and not overpowered by “decisions” made by the AI generator.
The upcoming third part will tackle whether it’s legal for AI models to train off copyrighted work. This guidance will impact all creatives with any amount of work that exists online and susceptible to AI scraping.
Shira Perlmutter, who heads the copyright office, summed up the second report:
Where that creativity is expressed through the use of AI systems, it continues to enjoy protection. An AI-assisted work could be copyrightable if an artist’s handiwork is perceptible.
How much or how little perceptible handiwork needs to be evident will be adjudicated on a case-by-case basis. This creates a grey area of who will determine it, and what biases might they have. Further, it’s possible that as an influx of items applying for copyright are sent to the office for deliberation, AI tools, rather than human insight, might be employed to determine whether an item should or shouldn’t be copywritten.
The report states that works solely generated with AI that do not show human intervention in the final product, cannnot be copywritten. That means, if a prompt was used, and the AI-generated result was not further manipulated by a human with creative intent, then it cannot be attributed to a human.
Foundational to this theory is that the copyright office instructed Google’s Gemini AI image generator with this specific prompt:
A professional photo, bespectacled cat in a robe reading the Sunday newspaper and smoking a pipe, foggy, wet, stormy, 70mm, cinematic, highly detailed wood, cinematic lighting, intricate, sharp focus, medium shot
It turns out the final product changes repeatedly despite the same prompt being used. Therefore, it was subject to the discretion of the AI system, and not decisions made by the person inserting the prompt.
Shira Perlmutter explains:
Extending protection to material whose expressive elements are determined by a machine... would undermine rather than further the constitutional goals of copyright.
One artist who may take issue with this governance is synthetic media artist, Jason Allen, creator of the hotly debated AI-generated image Théâtre D’opéra Spatial. He spends painstaking hours engineering the perfect prompt into Midjourney, and claims the AI-generated images are a result of his unique prompt.
The question is, can this effort be copywritten and attributed to a human author, and extend authorship to the image that results?
My grain of thought
The existing copyright laws, which date back to 1966, state that for a work to be attributed to a human, their influence must be evident in the final result. But they leave room for interpretation on exactly where in the process that influences should exist, and how much of it is required. This will both appease and displease artists, and we shall see how it plays out in public discourse, not to mention the courtroom.