So without rehashing all the details, there is a non-profit in Canada called CIPPIC, whose work I respect, and who has provided me with assistance in the past on an unrelated matter. CIPPIC does good public interest work at the intersection of law and technology.
They recently submitted a filing in federal court in Canada (a few months ago now) to ask for the correction of a copyright registration which was automatically granted by the Canadian Intellectual Property Office (CIPO) to a lawyer from India who used a style-transfer AI program to apply a Van Gogh Starry-Night-esque effect to an original photo they themselves took. As explained on Baker Botts site, and I believe quoting how the board of the US Copyright Office decided after the same person attempted to secure copyright in the US previous to Canada:
The Board explained:
“As Mr. Sahni admits, he provided three inputs to RAGHAV: a base image, a style image, and a ‘variable value determining the amount of style transfer.’ Sahni AI Description at 11. Because Mr. Sahni only provided these three inputs to RAHGAV, the RAGHAV app, not Mr. Sahni, was responsible for determining how to interpolate the base and style images in accordance with the style transfer value … Mr. Sahni did not control where [the Works] elements would be placed, whether they would appear in the output, and what colors would be applied to them—RAGHAV did.” (Office Letter p. 7)
Accordingly, the Board determined that the derivative work authorship was not the result of a human. Therefore, the Work was not registerable.
The Baker Botts site also shows the base image, the style image, and the output image. Sadly, the output image is not, in my opinion as a visual artist, actually any “good.” But that’s beside the point here.
This article on Norton Rose Fulbright gives a bit more context on Suryast and CIPPIC’s opposition to the copyright registration for the work, as filed with CIPO. Some snipped quotes for length:
The copyright registration lists RAGHAV Artificial Intelligence Painting App (RAGHAV) and Mr. Ankit Sahni as co-authors.
CIPPIC’s application challenges the copyright registration for Suryast and seeks expungement of the copyright, or in the alternative, removal of RAGHAV as a co-author. CIPPIC makes two main arguments: 1) Suryast does not meet the originality requirement for copyright; and 2) an AI system cannot be an “author” under the Copyright Act.
[…] However, CIPPIC submits that merely providing the inputs was a purely mechanical process and no human skill or judgment was used to produce Suryast. CIPPIC further contends that “author” in the Copyright Act only refers to a natural person (i.e., “human being”), and an AI system cannot exercise the common intent required for joint authorship.
My understanding, based on this York University Osgoode School of Law piece is that the actual Copyright Act of Canada does not explicitly define original/originality as concepts.
The question of what constitutes “original expression” in a work, though, required an answer from the Supreme Court, given that the Copyright Act does not itself define the terms (nor the term “original” by itself). As every student of copyright in this country also learns, in CCH, Chief Justice McLachlin wrote that:
“What is required to attract copyright protection in the expression of an idea is an exercise of skill and judgment.”
CIPPIC’s filing against the Suryast registration can be read in its entirety here. This is the most relevant part, though it is short and easy reading overall, so I recommend checking it out if interested:
- CIPPIC raises two alternative grounds for rectification:
a. the image lacks originality and so does not enjoy copyright at all; and
b. alternatively, a non-human cannot be an author under the Act.
i. The image is unoriginal
- The Suryast Registration should be expunged in its entirety pursuant to subsection 57(4)(b) of the Act because the image ought not to have been accepted
for registration at all: the Respondent has obtained a copyright registration in connection with an image in which copyright cannot subsist because it lacks
originality.
- The Respondent did not contribute sufficient skill and judgment in generating the image Suryast to warrant subsistence of copyright. The Respondent generated
the image through a purely mechanical exercise of data entry and algorithmic luck; its production is the result of no exercise of human skill or judgment.
That wording seems to refer back to the CCH case mentioned above.
Without having read or carefully studied all of the Copyright Act, I would have to agree that the computer program used to produce the image should not have been listed as a co-author, and the filing should be amended for that reason.
For transparency, I registered my AI Lore Books series with CIPO, but I registered the whole thing in my name alone, despite having used AI tools to produce elements of the contents.
But I think I disagree with CIPPIC’s assertion that no human skill or judgement was involved in producing this image. It’s probably useful here to go back to the longer quote from the full text of the CCH decision:
By skill, I mean the use of one’s knowledge, developed aptitude or practised ability in producing the work. By judgment, I mean the use of one’s capacity for discernment or ability to form an opinion or evaluation by comparing different possible options in producing the work. This exercise of skill and judgment will necessarily involve intellectual effort. The exercise of skill and judgment required to produce the work must not be so trivial that it could be characterized as a purely mechanical exercise. For example, any skill and judgment that might be involved in simply changing the font of a work to produce “another” work would be too trivial to merit copyright protection as an “original” work.
So:
Skill = developed aptitude, practiced ability.
Judgement = capacity for discernment, ability to form an opinion, evaluation by comparing different possible options.
I’ve seen mentioned in multiple places now that Sahni submitted at the USCO’s request a 17-page document detailing how he created the image and the technology involved. But I’m not able to currently locate it myself, though I’d like to see what it includes.
I’m going to go out on a limb here though and say that despite my not really “liking” the results of the style transfer, CIPPIC hasn’t made in that document much of any real case to explain why Suryast fails to demonstrate skill and judgement. They merely state that it is so by calling what Sahni did a “mechanical process.” But I would argue that what Sahni did rises well above the example cited in the CCH decision of simply changing a font.
Simply to know about AI/ML and style transfer on the part of the image’s (human) creator is the first demonstration that skill, aptitude, and ability may have significantly come into play in the image’s generation.
Without knowing the exact details of Sahni’s 17-page document detailing the creation of the image, it’s difficult to identify just how much skill and judgement was involved, but it seems to have absolutely been more than zero. According to this article, Sahni was the funder of the RAGHAV app which was built by an engineer named Raghav Gupta:
RAGHAV stands for robust artificially intelligent graphics and art visualizer, and is named after Raghav Gupta, a machine learning engineer who developed the app in 2019 in a funded project for Sahni.
An article on Holland & Knight adds a bit more from the USCO Suryast paper:
Footnotes 5 and 6 in the SURYAST decision discuss the lack of detailed evidence in the record as to how RAGHAV was designed and by whom (although RAGHAV was named for the engineer who developed the app for Mr. Sahni, Raghav Gupta). The Copyright Office, however, only had the vague description that RAGHAV was trained on a dataset of 14 million base images, called ImageNet, and then on another dataset of “content and style” images.
[…] If Mr. Sahni had designed RAGHAV and carefully selected its training materials, would that (in combination with taking the original photograph and selecting the style applied) constitute enough “creative control” for Mr. Sahni to assert authorship in the modification?
If the USCO is claiming they don’t have all the details about the RAGHAV software, then it makes me wonder what the contents of this mysteriously missing 17-page document actually were?
Regardless, even without that, it’s clear that judgement was involved, first in selection of the base and style image, and then in the decision of how much style to be transferred. All three of those components of the decision would have required the ability to form an opinion (“this is a good image to use”), and evaluate through comparison possible input images, style transfer settings to apply, and outputs received.
Never mind, of course, if Sahni did indeed hire someone to build this system for him based on criteria which he in part defined: all of which involved extraordinary skill and judgement to be applied.
So in conclusion, I would have to reiterate that I agree AI tools should not be listed on copyright registrations in Canada as co-authors because they do not constitute persons. But CIPPIC hasn’t proved – nor has anyone else so far to my satisfaction – that this work isn’t “original” under Canadian copyright law in that it lacks all skill and judgement, and is merely a mechanical process. As someone who uses AI tools constantly for creative work, it’s easy to say that framing of them doesn’t match the reality of using them: it’s a constant minefield, a battle, a struggle of skill and judgement, selection and direction, and so much more.
If for some weird reason this is interesting to people, then my longer submission to the USCO & CIPO public consultations on generative AI and copyright may also be. And I still think the current UK system for copyright of computer-generated works makes a lot more sense:
“The “author” of a “computer-generated work” (CGW) is defined as “the person by whom the arrangements necessary for the creation of the work are undertaken”. Protection lasts for 50 years from the date the work is made.”
My recent experiments with creating machines for the purposes of drawing has also made me fundamentally question the validity of the underlying assumptions of dismissing something based on it being the result of “merely mechanical” processes. The fact is that artists and creators can and do routinely manipulate mechanical processes in order to express a creative vision. And I’ll say it again: simply choosing one tool over another in order to express that creative vision shouldn’t invalidate it magically, when it would have been perfectly legitimate and accepted in another tool or media.
