Generative Art and Copyright — Part II
By Felix Yuen
In the previous article we introduced the generative art form. In this article we are going to look at some of the copyright problems that generative art brings about.
The (many) copyright issues
As generative art now often involves computers and AI, the question of whether the creator owns any copyright in the output created by the computers arises. One point of debate is that, with AI’s increasing autonomy, human intervention in the creation process has become minimal (the problem of lack of originality) or too remote that it is severed from the output. Some countries (for example, the UK and Canada) are looking to find an answer to this debate, and to revise the copyright law for clarification of the issue. But before legislative amendments are finally made, lawyers and judiciaries have already been trying to apply the existing outdated copyright regimes to the new technologies. This has created very fragmented and sometimes awkward results around the world (for example, possible copyright protection in the UK under the computer-generated work notion vs no protection in the US as confirmed by the Copyright Office Review Board earlier this year). This authorship topic has been extensively researched and explained — some analyses can be found here for the status quo in the EU and the UK.
This article instead discusses two other issues related to generative art: (i) the use of third party data and (ii) the role of programmers and their entitlement. The first issue concerns the fact that AI needs to be fed with materials, which often could be copyrighted, to learn and produce the outputs. The second issue concerns the important contribution of programmers in the creation of generative art and whether they warrant a share of the rights.
Feeding AI with copyright materials for machine learning — third party right and data mining issue
AI needs data to learn. For generative artists, the source of the works that they feed into the program could well be their own works, or copyright-free materials that they find on the internet. However, some artists may not be aware that the use of third parties’ copyright works to create derivative works (i.e. works based on existing works) may constitute copyright infringement, as the machine reproduces and adapts the original third party works, depending on the relevant jurisdiction. Here we will need to distinguish two approaches — the open-ended fair use copyright exception approach (e.g. in the US) and the close-ended copyright exception approach (e.g. in the EU and the UK).
The open-ended approach is in general more tolerant towards derivative works and in the US the most important factor is whether the output is transformative — that it adds something new with a further purpose or different character and does not substitute for the original use. On the other hand, the EU constantly finds the creation of derivative works without the permission of the owner of the original work infringing, except where the use falls within one of the few limited exceptions such as parody (which is often not applicable in case of generative art).
As AI programs continue to develop, we can imagine that there will soon be, if not already existing, software based on GAN or other deep learning models that are able to massively source data from the internet as training data instead of a limited pool. This is perhaps permissible in the US under the open-ended fair use regime, but not so in other close-ended copyright exception countries. In the latter group of countries, new specific exceptions such as data mining exceptions are needed to ensure legitimacy of such pooling of copyright materials. We see that some countries and regions such as Japan and the EU have introduced data mining exceptions into their copyright laws, but currently their scopes are often limited to scientific research and do not cover creation of art.
What is therefore important to generative artists is to refrain from using third party copyright materials or to obtain a licence from the copyright owners, unless they are so sure that they will be covered by the fair use exception (which unfortunately is uncertain by nature). They should also watch out for the data source of the software that they deploy for creating artworks and ensure that they are clear about the mechanism.
Programmers as principal contributors
Some artists take further steps to code and build the program themselves but many of them do not. It is simply not necessary for them to do so because there are existing codes and programs that would do a large part of the work for them. What they need to do is to choose the inputs to be fed in the system, adjust some parameters, and select the outputs that fit their artistic standards.
Where artists decide to use pre-existing codes and AI software, it is fair to say that the programmers contribute a substantial part of the creative process. This leads to the discussion of whether they deserve some rights in the outputs. Opponents would similiarise the case of artists and software, and that of writers and pen; and argue that software, like pens, is merely a tool. I would however differentiate the two cases for the reason that the programmers for generative art AI programs have a much larger impact on the outputs and also they often create the programs with a particular aim of generating a particular form of art. In comparison, the creators of generative art are not involved in the execution, which is instead done by algorithms preset by the programmers.
On this issue, most countries currently do not recognise the rights of programmers in the outputs created by the programs, mainly because the copyright laws of these countries do not recognise AI-generated work in the first place. The UK and several other common law countries such as New Zealand are rare exceptions as they do recognise computer-generated works and arguably endow some rights to programmers. This is due to the fact that the copyright legislation of these countries defines the author of computer-generated works as ‘the person by whom the arrangements necessary for the creation of the work are undertaken’ and there is no requirement that this person shall be the programmer or the artist / user of the program. Where the artist makes little effort to adjust the parameters of the program and simply deploys the program to generate outputs, it would seem more appropriate to consider the programmer as the one who makes the necessary arrangements and hence the author. As seen above, it should be noted that the UK legislation does not prescribe programmers as the authors and each case must be determined by its own merits.That said, currently the UK as well as New Zealand are looking at reforming their copyright laws and this question may possibly become a non-issue in a few years.
Taking one step back, in view of the possible legislative changes, we should perhaps ask the question whether programmers should be regarded as authors and own the rights in the outputs of the AI. I foresee two arguments against this. From the authorship perspective, programmers do not normally envisage the outcome of the AI at the time of its development, and so the relation between programmers and the output is not less remote than that between the artists and the output. From the financial reward perspective, programmers are already protected by the copyright law for the program they create and we may cast doubt on whether another layer of remuneration and protection is needed.
Further, a related question is whether programmers would then need to bear the liabilities of third party copyright infringement for the reason of the software’s ability to pool copyright materials for machine learning and to create derivative works, as mentioned in the previous section, if they are well considered as the authors of the final artworks.
As technology continues to rapidly develop, we could expect more questions to come.
The res ed cohort programme is an initiative by resolutio to help spread awareness on NFT rights. Learn about our cohort here. For updates, follow us on Twitter & LinkedIn and join our Discord Community.
