Peter Vaughan, Associate Professor, NLS https://www.ntu.ac.uk/staff-profiles/law/peter-vaughan
When it comes to images created by AI Generators much of the debate has focused on the training data (i.e. the collection of information used to train AI models) and the permissions required to use such data; the information going into the black box.
But what of the outputs? Do the images generated via AI generators attract protection through copyright law? Or should they be unprotected, free for anyone to use how they will as they have been created by a mysterious unknowable machine?
There has been an attempt in the US to grapple with this issue. The US Copyright Office in SR # 1-11743923581 Theatre D’Opera Spatial refused to award copyright protection to an image created wholly by an AI machine, which was then subsequently edited by the artist, unless the underlying AI image was disclaimed (although even then no final view was taken as the artist refused to consider this).
But what of the UK? How should its copyright law treat AI generated images?
To attract protection under copyright law in the UK, under Section 1(1) of the Copyright, Designs and Patents Act 1988, the output needs to be “original” and a “work”. The courts in THJ Systems confirmed original as being the “authors own intellectual creation”.
Are AI generated images created in this way “original”? Are they an “intellectual creation” of the author? Who is the author anyway? It can’t be the machine – they aren’t a person.
The answer to the latter question is fairly simple. A copyright work must have an author, and an author needs to be a person. Section 9(3) Copyright Designs and Patents Act 1988 tells us that the author of a computer generated work is “the person by whom the arrangements necessary for the creation of the work are undertaken”. This provides a convenient answer to who the author is, likely being the person who inputted the prompts into the AI generator, thereby sidestepping any questions around machine based ownership, and allowing for focus on originality.
We must therefore turn to “originality”. This is the key requirement for a work to attract copyright protection. The answer, it seems, comes down to the prompts (the text based commands used to direct the machine).
For a simple prompt, AI Image generators are not much more than a random image generator – ask for an image and cross your fingers. If the prompt is limited to just asking for an image at a general level of abstraction, then there is no real control over the end image. “Creative choice” (if I may be forgiven a personification) has been delegated to the machine.
What is the position when the commands are more complex? Where the direction given to the machine is more directive and akin to a creative piece of writing in itself this could meet the threshold of ‘creative choice’. At this level, a lot of direction is provided to the AI generator and the situation is more akin to that of an architect instructing a draughtsperson (those lucky individuals tasked with creating the technical drawings from sketches and measurements). In such a situation the courts have recognised that the architect is at least an author. Why, then, not in the situation where the draughter, acting on detailed instructions, is a machine?
As anyone who has played with AI image generators will know, there is a skill to creating the right prompt. Wording matters, as does the level of detail given. There can also be an element of trial and error. It is rare, if seeking to realise a specific image, to get it right on a first go. Prompts are refined as the image gets closer to the desired outcome.
This step suggests skill and judgement is required. Under the old copyright standard, this may well have been enough to attribute copyright to the image. But this type of prompt development requires creativity; it involves the expression of an imaginative aim and is more than just a formulaic composition. In involving creativity, it meets the required threshold of originality.
The approach of the US Copyright Office was therefore too restrictive and conservative in its approach to an emerging art form. For those who want it, at least in the UK, there is a path to granting protection for such AI generated images where the underlying creativity of the author as seen in the prompts is reflected in the AI generated image. That such an image might then be sharpened and tweaked using more traditional tools should not preclude protection. In this way, AI Images are to be seen not as created by a mysterious black box but by an artist using a tool of the trade. It might be that the scope of protection needs to be considered but this should not mean that the British courts use originality as a hurdle to avoid more difficult questions.
Further Reading:
Comentarios