AI Authorship and Copyright Protection

1.0 INTRODUCTION

Contemporarily, in a European context, approaches towards artificial intelligence (herein ‘AI’) for the purposes of technological advancement and innovation have been rather favourable. This has been indicated by the establishment of the ‘European AI Alliance’ in 2018, facilitating open policy dialogues on recent developments within AI and more. Whilst the ongoing effects of AI have already been witnessed, such as facial and voice recognition sensors in devices, autonomous vehicles in the form of self-driving cars; and as of 2022, the infamous ‘ChatGPT’ AI chatbot – resembling a calculator of sorts with an ability to generate conversations with humans. With the development of artificial narrow intelligence, systems can produce artworks and other creations seamlessly without the labour of a humanistic, physical element.

Fundamentally, the challenge in assessing intellectual property rights (referred to as ‘IPRs’ herein) comes with the issue that AI does not smoothly fit within the mould and legal system governing IPRs, due to traditional legal concepts constructed at a time pre-dating the advent of AI. At the same time, in the case of AI-generated artistic works, the governability of IPRs becomes rather nebulous, considering the non-excludability and non-rivalry of art in itself – free for all.

The overarching question this article aims to answer is whether, from an IP-rationale perspective, it is justifiable to protect AI-assisted/AI-generated creative and technical works through IP rights, with a focus on copyright protection. This will be performed by addressing the distinction between AI-generated work and AI-assisted work. Secondly, the assessment of the AI-generated work in question, namely the revolutionary Edmond de Belamy painting, will be done in relation to the subject matter of copyrights, and the protection(s) they offer. The last section of this paper will thus focus on whether theories of IP can justify their protection.

2.0 AI-assisted work versus AI-generated works

It can be said that AI-assisted works refers to outputs that must be within the realm of literary, scientific and artistic domain – whatever mode or form of its expression – as stipulated in Article 2 of the Berne Convention. Moreover, the AI-assisted output must be the result of human intellectual effort, with the Court of Justice of the European Union (CJEU) clarifying this in the Painer case, whereby works of authorship can be administered with the assistance of a machine or device. As such, a hybrid is made between human input through, for example, gathering data and developing software, and actual AI technology.

On the other hand, AI-generated output is made without any human intervention, referring to algorithms that create new outputs – namely photos, videos, codes, texts, etcetera. As such, AI-generated works produce information through trained programmes, most commonly known as generative adversarial networks (machine learning (ML) algorithm). This system arguably scarcely involves human intervention.

We can apply these concepts of AI-assisted outputs and AI-generated outputs in the case of the Edmond de Belamy painting (the first original work of art produced using AI) since the enterprise ‘Obvious, AI & Art’ trained an algorithm on a set of 15, 000 portraits from WikiArt, it can be said that such a painting is a product of an AI-generated output. This is further supported by the indication that there was minimal human intervention, of which the painting was, rather, produced by the AI itself via an algorithm. It is important to note that the human input lacks in this case due to the weight of AI input weighing larger than user input, such as in the form of compilation of data or human instruction.

3.0 Subject-matter: copyright

With regard to copyright, as we are dealing with authorship, to which the Berne Convention touches upon, it may be argued that ‘author’ is not clearly defined. However, if the author’s name is indicated, they shall be regarded the author of a literary or artistic work – opening the doors to both natural and legal persons. This definition stretches insofar that the author of a computer program shall be designated as a right-holder. In the case of the Edmond de Belamy painting as both an artistic work and AI-generated output, authorship would thus be given to the art produced by AI technology, extracting data from the thousands of portraits on WikiArt. As such, since copyrights are unregistered rights, they exist from the moment of creation and are thus assigned by default. Presently, when considering the legal frameworks governing IP rules, copyright protection exists insofar that authors must be natural persons, highlighting the humanistic element as shown in Article 7 of the Berne Convention prescribing the length of protection granted to an ‘author’ fifty years after ‘his’ death. This poses complications considering the reproductive nature of AI technology.

Regarding protectability for copyright, the requirement of originality is critical. Originality can be defined as the authors own intellectual creation – reflecting a personal element in the work through free and creative choices. Since the ‘author’ in the case of the Belamy painting is the AI entity, thus possessing its own intellectual creation of sorts (through the pre-programmed, trained algorithm on 15,000 portraits), it can be said that the requirement of originality is demonstrated. Despite the origin of the artwork, or the manufacturer behind it, the tangible expression of the idea – in this case the blend of 15,000 portraits as the source which thus manifests into a painting on a canvas – is protected.

In terms of the enjoyment of rights, in the case of the painting, it can be said that Articles 9 and 11, respectively, on right of reproduction and public recitation of the Berne Convention, are relevant. Such rights would, in a superficial sense, be enjoyed by the AI, however, the lack of human intervention would make distribution difficult.

Overall, although AI offers a step forward into technological developments and its many advantages, it is clear that States remain hesitant in AI authorship due to the key, human role for the protection of works and, in general, AI-generated creations, as evidenced by the U.S. Patent and Trademark Office’s refusal to accept an individual’s patent-related applications, for example.

4.0 Justifying IP protection? A brief overview

The theories that will be covered in this article mainly include the utilitarian theories (or deontological justification theories) of intellectual property. Based on the idea of balancing the flow of ideas and incentivising investment, a balance must be ensured to encourage the flow and free access to knowledge whilst maintaining sufficient investment opportunities (or the incentive to do so).

According to John Locke’s labour theory, people are entitled to their property rights due to the labour they have incurred (onto something – assuming a physical element exists), and thus should earn the ‘fruits’ of that labour.  With this idea of carrying labour toward intellectual work, in the case of the Belamy painting, the AI would be entitled property rights as carrying out intellectual work, in addition to the data generated through the 15,000 paintings originally sorted through ‘Obvious AI & Art’ as a legal person (and therefore right-holder) the result of which an intellectual good is finalised. Such argumentation follows the premise that since the AI entity is responsible for the creative output in the production of the painting, ‘it’ is entitled to (intellectual) property rights. This theory is arguably one-dimensional, ruling out the teams of human individuals that have fed the algorithm in the first place.

Secondly, the personality theory originating from Georg Hegel outlines that generating something and making it accessible to the general public is an expression of personality. This theory essentially connects the concept of personality to the intellectual work which embodies its creator’s personality or will – containing social and economic aspects through the security of being self-reliant by owning something with dignity and honour. It can be argued that with the painting, the AI, which based it on a combination of other paintings, attaches its own uniqueness in contributing to the arts, simply through the avenue of AI as the engineer behind the painting. Essentially, the originator of the work deserves recognition due to the originality behind the author’s ideas. This moral right conflicts with AI as an impersonal automaton that is programmed with an individual’s, or group of individuals, ideas.

Thirdly, Jeremy Bentham’s reward theory supports the idea that lawmakers should strive towards the maximalization of happiness (arguably the strongest of utilitarian theories) and social welfare. The reward of copyright protection, for instance, is that it is useful to society in fostering the creation of literary and artistic works. Applying this to the Belamy painting, it can be concluded that such a creation would benefit society in being able to use it freely without curtailing public enjoyment, whilst maintain exclusivity to stimulate innovation.

Conclusively, rather than adopting a single, clear-cut approach (such as the property centred approach giving sole legal personhood to non-humans, or property-last approach attaching no IPRs to non-humans), all theories put forth present valuable arguments in support of the justification of copyright protection whilst acknowledging potential drawbacks. The possibility of a middle-ground, steering away from the traditionalist theories of (intellectual) property law, such as the modernist school of thought, might be optimal – allowing for the human presence in authorship as the vessel who manages the arrangements for the creation of a work.