A LEGAL APPROACH TO WHETHER AI

GENERATED CONTENT SHOULD BE

PROTECTED UNDER COPYRIGHT
By

1839935

Submitted in partial fulfilment of the requirements for the degree of

Master of Laws by Coursework and Research Report

at the University of the Witwatersrand, Johannesburg

Date: 2 October 2023

1
DECLARATION
I declare that this report is my own, unaided work. It is submitted in partial fulfilment of the
requirements for the degree of Master of Laws in the field of Information and Communications
Law at the University of the Witwatersrand, Johannesburg. It has not been submitted before
for any other degree or examination in any other university.

I have submitted my final Research Report through Turnitin and have attached the report to
my submission.

Word Count: 10,399

_________________________________
1839935

1 October 2023

2
ABSTRACT
Currently, there is no clear answer for whether AI-generated content should be protected under
copyright law in South Africa and if so, who is the author and who owns the copyright. As AI
is growing more advanced and widely used, the potential for confusion grows and thus clarity
on the law’s position is important. In this paper, I will examine how AI generates works and
compare this to the purpose of copyright law. I will then determine whether granting AI-
generated works copyright protection aligns with the purpose of copyright law.

Further, I will conduct an analysis on how different countries have dealt with the question of
whether AI-generated content should be protected under copyright. This will help pinpoint the
factors to consider when answering this question such as the requirement of human authorship,
legal or juristic person ownership of the copyright and whether the works are copyrightable. It
is important to consider how South Africa’s legal system should approach issues surrounding
AI-generated content and copyright.

In examining the South African legal position on this question, I will conduct an analysis on
the approaches taken by the US, the UK, Germany, Australia, China, and South Africa in
relation to AI-generated works and copyright. This range of countries will allow for a greater
understanding of the issues, complexities, and factors to consider while answering this
question.

3
Table of Contents
DECLARATION ......................................................................................................... 2
ABSTRACT ............................................................................................................... 3
I. INTRODUCTION ................................................................................................. 5
II. What are AI-generated works? ...................................................................... 6
III. What is the purpose of copyright? ................................................................ 7
(a) What are the justifications of copyright law? ............................................................... 8

(b) Who are the subjects of copyright law?...................................................................... 10

(c) What works are eligible for copyright protection in South Africa? ............................... 12
(i) Payen v Bovic ....................................................................................................................... 13

(d) AI and the Copyright Amendment Bill ...................................................................... 14

IV. Does copyright law extend to AI-generated works? – A comparative law

exercise. .................................................................................................................. 16
(a) Existing precedents around the world........................................................................ 16
(i) The US................................................................................................................................. 16
(ii) The UK ................................................................................................................................ 19
(iii) Germany .............................................................................................................................. 21
(iv) Australia .............................................................................................................................. 23
(v) China................................................................................................................................... 24

V. Should AI-generated works be protectable under copyright law in South

Africa? ..................................................................................................................... 26
VI. CONCLUSION ................................................................................................ 30
BIBLIOGRAPHY ...................................................................................................... 32

4
I. INTRODUCTION
Artificial Intelligence (AI) can be defined as computer systems performing tasks using pattern
recognition to make decisions and learn from experience.1 AI has many differing definitions
as it is ever evolving.2 The concept of intelligence itself is vague so AI researchers mainly refer
to the notion of rationality. Rationality is the ability of AI to choose the best path towards a
goal.3 AI is currently mostly assisted intelligence rather than truly autonomous intelligence as
AI works as a tool to aid humans in their tasks.4

AI-generated content’s constant development and wide usage in our societies has led to
questions about whether these works should be protectable under copyright law.5 In South
Africa, our courts have yet to deal with this question. Thus, it is useful to engage in a
comparative analysis of the current approaches in the US, UK, Germany, Australia, and China.
This can provide a deeper understanding of what factors to consider, the possible issues as well
as how best to give effect to the purpose of copyright law.

I will analyse how AI generates content as well as the purpose of copyright. I will then
determine if AI-generated works fit within the parameters of copyright protection based on the
justifications of copyright. If the purpose of copyright is to incentivise creation through
economic and moral rights, then AI would not require copyright protection for its works. AI is
not motivated through financial or moral means. It will continue to create as it was coded to do
so.

Further, I will examine how different countries approach the question of copyright in AI-
generated works.6 This will provide clarity on the factors South Africa should consider when
dealing with AI-generated works and copyright.7 It may also aid in answering whether there

1
H Sheikh, C Prins, E Schrijvers ‘Artificial Intelligence: Definition and Background’ In: Mission AI. Research
for Policy (2023) s1.
2
Ibid at s2.
3
European Commission’s High-Level Expert Group on Artificial Intelligence ‘A definition of AI: Main
capabilities and scientific disciplines’ available at
https://ec.europa.eu/futurium/en/system/files/ged/ai_hleg_definition_of_ai_18_december_1.pdf accessed on 20
July 2023 1.
4
H Hassani, ES Silva, S Unger, M TajMazinani, S Mac Feely ‘Artificial Intelligence (AI) or Intelligence
Augmentation (IA): What Is the Future?’ AI (2020) 1(2):143-155.
5
A Zhuk ‘Navigating the legal landscape of AI copyright: a comparative analysis of EU, US, and Chinese
approaches’ AI Ethics (2023) at 1-2.
6
Section 223 of the Constitution of the Republic of South Africa, 1996.
7
Ibid.

5
should be any changes to the Copyright Amendment Bill to make the law clearer and avoid
future issues.

II. WHAT ARE AI-GENERATED WORKS?

To answer whether AI-generated works should be protectable under copyright, the nature of
how AI generates works must first be ascertained. AI Engineering is an extension of software
engineering.8 AI developers write code setting parameters for the AI. The AI proceeds to
generate works through a neural network which performs pattern recognition.9 Al
developers/engineers are hired by companies and paid as an incentive to create AI.

Handcrafted Knowledge Systems (“HKS”) were the first “AI”. HKS’s are AI which use long
series of programmer made inputs and outputs in a rule-based software system to function (if
‘a’ happens then do ‘b’).10 This was a codification of human knowledge. The computer was
not learning for itself, it was following a script set out by the developers.11

Machine learning AI systems of today work differently. Their knowledge is not taken directly
from human programmers’ hand coded rules.12 Rather the programmers provide the AI with
training data and then the AI must learn what the correct action is through trial and error using
pattern recognition.13 The human programmers aid in the learning process by guiding the AI
through the choice of datasets, what algorithms to run and in troubleshooting.14

AI-generated images usually use Generative Adversarial Networks (GANs) to produce images
with modifiers or improvements to increase realism and originality.15 GANs use deep learning
(three or more neural networks) and a convolutional neural network (CNN) to analyse visual

8
J Bosch, H.H. Olsson, & I Crnkovic ‘Engineering AI Systems: A Research Agenda’ (2021) Artificial
Intelligence Paradigms for Smart Cyber-Physical Systems 1-19.
9
P Boucher ‘Artificial intelligence: How does it work, why does it matter, and what can we do about it?’ (2020)
EPRS | European Parliamentary Research Service 3.
10
J.G. Hoffer, A.B. Ofner, F.M. Rohrhofer et al. ‘Theory-inspired machine learning—towards a synergy
between knowledge and data’ (2022) Weld World 66, 1291–1304; J Launchbury ‘A DARPA Perspective on
Artificial Intelligence’ available at https://www.darpa.mil/attachments/AIFull.pdf accessed on 5 April 2023 5-6.
11
Ibid at 7.
12
Boucher op cit note 9 at 3.
13
Ibid at 3.
14
G Allen ‘Understanding AI Technology’ Joint Artificial Intelligence Center (2020) at 7-8.
15
E Cetinic & J She ‘Understanding and Creating Art with AI: Review and Outlook’ (2022) ACM Transactions
on Multimedia Computing, Communications, and Applications 10.

6
images. GANs sort through large data sets, find patterns and then generate a new example to
blend in with the original data set.16

GANs have two sub-models namely the generator model (trained to generate new realistic
images) and the discriminator model which deems the new images as fake (AI-generated) or
real (from the original dataset). The aim is for the generator model to produce such realistically
original images that the discriminator model as well as humans cannot tell that it was AI-
generated.17

AI-generated writings are created using Large Language Models (LLMs). LLMs are developed
using deep learning models containing a billion parameters. They are trained in natural
language processing through large quantities of unlabelled data and human-written text.18 The
number of datasets as well as the fine tuning done by software engineers allows the AI to learn
to generate its own writings.19 This is done using prompts from humans as the AI’s instructions.
AI is usually a co-creator of content rather than the sole creator.20 The possible copyright issue
which may arise for both GANs and LLM AI-generated works is that the training datasets can
contain copyright protected works leading to concerns regarding copyright infringement.

AI generates work using human input to guide it. AI does not generate works without human
intervention yet. So, AI-generated works that exist today could be referred to rather as AI-aided
works. This is important to note as it becomes apparent in the next sections that human
authorship is necessary for copyright protection. However, how much human intervention is
needed to meet the authorship and originality requirements within copyright law needs to be
interrogated.

III. WHAT IS THE PURPOSE OF COPYRIGHT?

To assess whether AI-generated content should be protected under copyright law I will first
examine the purpose of copyright. This will allow me to determine whether having AI-

16
Ibid at 9-10.
17
Ibid at 9-10.
18
I Alberts, L Mercolli, T Pyka et al. ‘Large language models (LLM) and ChatGPT: what will the impact on
nuclear medicine be?’ (2023) Eur J Nucl Med Mol Imaging 50, 1549–1552.
19
Ibid.
20
D Yang, Y Zhou, Z Zhang, T. J-J Li, & R LC ‘AI as an Active Writer: Interaction strategies with generated
text in human-AI collaborative fiction writing’. In A. Smith-Renner, & O. Amir (eds.), Joint Proceedings of the
IUI 2022 Workshops: APEx-UI, HAI-GEN, HEALTHI, HUMANIZE, TExSS, SOCIALIZE (2022) 56-65.

7
generated content protected under copyright law aligns with the purpose of copyright and thus
whether it should be protected.

(a) What are the justifications of copyright law?

I will discuss five theories of copyright justification. The appropriation theory which includes
the labour-based theory (including the rewards theory) and the personality theory, the
utilitarian theory, and finally the economic theory.21 Appropriation theories are based on the
idea that when one takes possession of an idea or a way of material production that they become
the lawful owner thereof.22 This theory has a focus on the creators’ authorial contributions.
They are creator-centric and use the creator’s self-actualization or labour to justify their right
to ownership.23

Further, the labour theory, as set out by John Locke, says that people have property rights over
the creations resulting from their own labour.24 The Lockean labour theory allows for such
works to be available to the commons (the public) as raw material on which to add their own
labour.25 This would support how AI generates creative works as it uses writings and images
in the online sphere to generate its own work. Thus, the AI is adding its own labour to raw
materials, it would then be the owner of the resultant work. However, the Lockean labour
theory is aimed at people as it is based on their individual natural right to property over the
products of their labour and AI is not a person.26

In addition, the rewards theory posits that copyright awards ownership (exclusive legal rights)
over original creative works. This allows for the monetization of creative works. This then
incentivises people to keep putting in their labour to create, to benefit society.27 Copyright
rewards creators for creating social utility.28 AI itself does not need to earn money or be
awarded legal rights to incentivize it to continue to generate content. AI does what it is coded
to do. The people and companies responsible for the coding of the AI are already incentivised

21
M Longan ‘A System Out of Balance: A Critical Analysis of Philosophical Justifications for Copyright Law
through the Lenz of Users’ Rights’ (2022) Michigan Journal of Law Reform 56.
22
Ibid 4-5.
23
Ibid 4-5.
24
T Aplin & J Davis ‘Copyright I: history, justifications, sources of law, and subsistence’ in Intellectual
Property (2022) 4-7.
25
Ibid.
26
Ibid 5-6.
27
Longan op cit note 21.
28
M Du Bois ‘Justificatory Theories for Intellectual Property Viewed through the Constitutional Prism’ (2018)
PER / PELJ 21,19-21.

8
to continue creating and improving the AI as they are paid to do so.29 Thus, the rewards which
come from copyright protection are unnecessary for both AI itself as well as AI creators to
continue to produce works. They are already being rewarded for creating social utility.

The second appropriation theory, the Hegelian personality theory, would not support allowing
AI-generated works protection under copyright. It posits that creative works are tied to the
creator’s well-being and sense of self. Hegel saw intellectual property (“IP”) as a mechanism
for protecting creators’ self-expression and dignity.30 Personality theory supports copyright law
as the protection of the creators’ moral rights.31 AI does not create works as a means of self-
expression, nor is it able to claim moral rights as it has no legal personality. Machines are not
afforded dignity under s10 of the Constitution of the Republic of South Africa as the heading
states “human dignity”.32 However, the creators and users of the AI are human beings and are
afforded the right to dignity. Thus, personality theory would only allow the creators or users
of the AI to claim copyright protection.

The economic theory posits that the high costs for the initial creation and the marginal
distribution costs of IP result in a market failure. Where creators cannot recuperate the money
spent on production and distribution, they will stop creating resulting in inefficient resource
allocation. Copyright law can correct this by creating an economic incentive for creators.33 IP
is non-rivalrous as when one person uses the product, others may still use it without there being
a reduction of the product. The creation of a property right through copyright law creates a
limited monopoly and provides an incentive for creators to create.34 AI is not in need of
financial or economic incentivisation in the same way as their human counterparts. The AI’s
human creators do not require further incentives to create as they are already paid to do so
within their jobs. Thus, the economic theory would only justify the companies and humans
behind the AI to claim for copyright of the AI algorithm itself.

The Utilitarian theory uses the philosophy that the best action results in the greatest net
happiness for the most people with the least suffering. When applied to copyright law, this
would skew towards the interests of the users/consumers as there will almost always be more

29
E Bonadio & L McDonagh ‘Artificial Intelligence as Producer and Consumer of Copyright Works:
Evaluating the Consequences of Algorithmic Creativity’ (2020) Intellectual Property Quarterly 6.
30
Du Bois op cit note 28 at 19-21.
31
Ibid.
32
Section 10 of the Constitution of the Republic of South Africa, 1996.
33
Du Bois op cit note 28 at 27-28.
34
Ibid at 27-28.

9
users/consumers than creators for a product.35 A utilitarian construct may allow for free reign
of the public over creative works. However, without any protections through copyright there
would be a decrease in creative ideas and works and thus a decrease in the happiness of the
majority.36 Utilitarian theory suggests that there be the least number of protections able to
produce the most happiness. This is done by having enough protections to promote creation
but not too many to stifle the enjoyment of such creations by the public.37 The current South
African copyright law perhaps offers too wide of a protective net over creative works to
produce this result and thus are not currently utilitarian.

For the purposes of this paper, I accept that the purpose of copyright is to work as an incentive
to encourage the creation of works by humans.38 Copyright incentivises creation through the
protection of moral rights and the granting of exclusive legal rights which results in financial
and economic incentives. None of these incentives are useful when applied to AI as it creates
as it is coded to do so. Thus, the copyright justifications do not support AI-generated works
receiving copyright protection.

(b) Who are the subjects of copyright law?

The subjects of copyright law are the person or legal entity who owns the copyright and the
users of the copyrighted materials. There are also intermediaries such as publishers who can
complete a contract with the author to take assignment of the copyright. The contract allows
the publisher to gain ownership (partly or wholly) over the copyright protected work.39 This
shows that the author of the copyright may not always be the owner of the copyright. The same
is also true with companies. Where employees’ author a creation in the course and scope of
their employment under a service or apprenticeship contract, the employer maintains
ownership over the copyright thereof.40 The copyright owner of commissioned works is the
person who paid for the work to be commissioned.41 A user-rights approach to copyright
considers the permissible uses of copyrighted materials as the user’s rights. It emphasises the

35
Longan op cit note 21 at 20-24.
36
Ibid at 20-24.
37
Ibid.
38
Ibid at 29.
39
Section 22 of the Copyright Act 98 of 1978.
40
Ibid at s21(1)(d); King v SA Weather Service 2009 (3) SA 13 (SCA).
41
Ibid s21(1)(c).

10
user’s role as partner to the creators in achieving the objectives of copyright.42 Users are not
just making use of copyrighted material. Users are active participants in the promotion of the
creation, dissemination, and use of copyrighted works.43

South Africa is party to the Berne Convention for the Protection of Literary and Artistic Works
(“the Berne Convention”).44 The Berne Convention seeks to protect the rights of authors and
vest in them control over the use of their works.45 It provides authors with economic and moral
rights,46 neither of which is useful in incentivising AI to create as discussed above. While there
is no definition of ‘author’ within the Berne Convention, the wording suggests that it refers to
human authors though the final determination of authorship is left to national law.47 AI would
not be considered an author within the Berne Convention under Art. 2. Berne members then
have no obligation to protect purely AI-generated works even where the origin countries have
chosen to protect them under copyright.48

South Africa is party to the Marrakesh Agreement which established the World Trade
Organization (“WTO”). South Africa as a member of WTO is automatically bound by the
TRIPS agreement.49 The goals of the TRIPS Agreement are to reduce impediments to
international trade, to promote protection of intellectual property rights (“IPRs”) and to ensure
enforcement procedures for IPRs do not become barriers to trade.50 Protection and enforcement
of IPRs should contribute to the promotion of technological dissemination and innovation to
the users and producers mutual advantage.51 As AI will continue to be improved upon by
technology company employees, technological innovation will be promoted without the need
to extend copyright protection to AI-generated works. Furthermore, if AI-generated works
were to be unprotectable by copyright then those works would become a part of the public

42
N Elkin-Koren ‘Copyright in a Digital Ecosystem: A User-Rights Approach’ in R Okediji (ed) Copyright in
an age of Limitations and Exceptions (2015) 2-3.
43
Ibid at 2-3.
44
WIPO ‘WIPO-Administered Treaties: Contracting Parties Berne Convention’ available at
https://www.wipo.int/wipolex/en/treaties/ShowResults?search_what=C&treaty_id=15 accessed on 15 April
2023.
45
Berne Convention for the Protection of Literary and Artistic Works (as amended on September 28, 1979) at
Article 1.
46
Ibid at Article 6.
47
JC Ginsburg ‘People Not Machines: Authorship and What It Means in the Berne Convention’ (2018) IIC 49,
131–135 at 135.
48
Ibid.
49
TRIPS: Agreement on Trade-Related Aspects of Intellectual Property Rights available at
https://www.wto.org/english/docs_e/legal_e/27-trips_01_e.htm accessed on 10 April 2023.
50
Ibid at Preamble.
51
Ibid at Article 7.

11
domain. This would help balance the rights of users and producers by adding to the information
commons and promoting further creation.52

(c) What works are eligible for copyright protection in South Africa?
The Copyright Act 98 of 1978 provides protection for nine categories of works namely literary
works, artistic works, musical works, cinematograph films, broadcasts, sound recordings and
programme-carrying signals as well as computer programs and published editions.53 Electronic
databases are included in literary works and underlying code is included in computer programs.
Thus, the code underlying an AI program would be protected under copyright law.54

There are four requirements that creative works within the categories must meet to enjoy
copyright protection. These are: originality, existence in a material form, the author must be a
‘qualified person’ as defined under s3(1) of the Copyright Act and publication.55 Originality
requires that the creator or author used skill and labour in the creation process. South African
courts do not look for inventiveness or creativity. They consider the degree or amount of labour,
judgement and skill applied during creation.56 The skill, labour and judgement must be
substantial rather than minimal.57 How much skill or labour is necessary for originality is
decided on the facts of each individual case.58 Time and effort are not enough.59

There is no copyright in ideas, existence in a material form is required.60 A qualified person is

either a South African citizen or a person who is resident or domiciled61 in South Africa.62
Further, a qualified juristic person is a company incorporated under South African laws.63 AI
is not a natural person nor a juristic person, AI does not have legal personality and cannot be

52
AJ van der Walt & M du Bois ‘The importance of the commons in the context of
intellectual property’ (2013) 1 Stell LR 32, 44-46.
53
Copyright Act op cit note 39 at s2(1).
54
Ibid at s2(1).
55
Ibid at s2(2).
56
Kalamazoo Division (Pty) Ltd v Gay 1978 (2) SA 184 (C) para 192.
57
Haupt t/a Softcopy v Brewers Marketing Intelligence (Pty) Ltd 2006 (4) SA 458 (SCA) para 35.
58
Klep Valves (Pty) Ltd v Saunders Valve Co Ltd 1987 (2) SA 1 (A) at 22H – 23A.
59
Moneyweb (Pty) Ltd v Media 24 Ltd & another 2016 3 All SA 193 (GJ); 2016 (4) SA 591 (GJ).
60
W Alberts “Copyright in ideas: A lesson from the ivory tower” (2008) Juta’s Business Law 48.
61
LexisNexis & Squire Sanders ‘International aspects of copyright’ available at
https://www.squirepattonboggs.com/~/media/files/insights/publications/2012/10/international-aspects-of-
copyright/files/international-aspects-of-copyright-practice%20note/fileattachment/international-aspects-of-
copyright-practice-note.pdf accessed on 8 April 2023: The Berne Convention provides a system of reciprocity
allowing foreign copyright owners to be treated as a national of the place where an infringement took place so
they can claim copyright protection there. This is known as the principle of national treatment.
62
Copyright Act op cit note 39 at s3(1).
63
Ibid at s3(1).

12
considered a qualified person. Finally, publication involves copies of the relevant work being
issued to the public with the copyright owners’ permission in a quantity that can meet the needs
of the public.64

A qualified person is necessarily a human or a juristic person, there is no mention of machines

or computer systems being qualified people.65 Further, originality requires the author to have
used a substantial degree of skill, labour, and judgement in the creation of the works. This does
not lend itself to providing AI-generated works created without a qualified persons labour and
skill with copyright protection. AI works generated with enough human skill and labour
involved would be able to be protected through copyright. What ‘enough human skill and
labour’ means is determined by the court in a fact-based enquiry on a case-by-case basis. It has
yet to be decided which prompts amount to having been written with substantial skill and
labour. The Act states the author of an artistic, dramatic, literary work or computer program
that is computer-generated would be the person who made the necessary arrangements for the
creation.66 The extent of human intervention needed to meet the necessary arrangement
requirement has yet to be interrogated.

(i) Payen v Bovic

In analysing how the Copyright Act and our court system should deal with the question of
copyright protection of AI-generated works the Payen v Bovic case is useful. In Payen v Bovic,
both parties were juristic persons. Payen’s code catalogue was used without permission by
Bovic to identify different types of gaskets. Payen claimed that Bovic had infringed their
copyright.67 Payen developed this catalogue through a long process involving manual effort
and skill on the part of Payen’s human employees. The Court decided that there was a
significant amount of skill, labour and knowledge which went into the code catalogue and as
it was in a material form, it counted as an original literary work.68

However, Bovic argued that the computer system itself was the thing which authored the code
catalogue. They argued that a human author is required for copyright protection. Thus, there
can be no copyright protecting an AI-generated catalogue. While the computer program used

64
Copyright Act op cit note 39 at s42.
65
Ibid at s3(1).
66
Ibid at s1.
67
Payen Components SA Ltd v Bovic Gaskets CC 1995 (4) SA 441 (A) 4-12.
68
Ibid 9-13.

13
by Payen is protectable by copyright as a literary work, the catalogue it produces is a different
work entirely.69

Further, the Court considered the Express Newspapers70 case from the UK. In that case, Mr
Erthel who was contracted by Express Newspapers to work on a system of numbers using a
computer system for a random number newspaper competition. It was shown that he had used
considerable skill and labour in the creation of the system.71 The judge stated that arguing the
computer was the author would be akin to arguing a pen is the author rather than the man who
wielded the pen. The computer system was being used as a tool.72 It was acknowledged that
giving the programmer of the computer the authorship of the computers work would be
misleading and inconvenient.73

The court defined the differences between computer-aided and computer-generated works.
Computer-aided work is where the computer is used as a tool to assist the human author in
creation and thus the usual authorship considerations apply. In computer-generated works there
is minimal human involvement, and the creation process is mostly autonomous.74 This was a
case of computer-aided work not computer-generated work and thus there was a human author
and copyright protection could be granted.75

Moreover, in the Payen case the court decided that the human workers at Payen had put enough
skill and labour into the development of the code catalogue program (human authors). It was a
case of computer-assisted literary works rather than computer-generated and thus it was
afforded copyright protection. Payen had copyright ownership over the works.76 The South
African court in this case followed the UK’s approach to the copyright protection of AI-
generated works.

(d) AI and the Copyright Amendment Bill

The Copyright Amendment Bill (B13-2017) (“CAB”) aims to amend the Copyright Act, 1978
to promote balanced accessibility for users, producers, and consumers; including digital

69
Ibid.
70
Express Newspapers PLC v Liverpool Daily Post & Echo PLC and others 1985 1 WLR 1089 (Ch): 1985 FSR
306.
71
Ibid.
72
Ibid.
73
Ibid.
74
Payen supra note 67 at 450D-G.
75
Ibid at 14.
76
Ibid 18-19.

14
advancements to empower citizens.77 It also aims to address licensing of copyrighted works
for commissions to facilitate commercial exploits by the licensee.78

Copyright was developed as an incentivising system to encourage human creation. It does so

by allowing authors to earn money from their works and to protect their personal expression.
AI is not human and does not need these incentives.79 It will create so long as human input is
given.80 Humans will continue to provide the AI inputs as some are paid to do so and others
use AI as a creative tool without requiring ownership over the works produced. Humans then
do not need copyright protection over the works AI produces. Thus, it is not necessary to extend
the definition of ‘author’ to include the person who made the necessary arrangements for a
work generated by a computer.

Further, there are concerns regarding the training of AI using copyright protected material.81
AI mimics (imitates) or amalgamates (combines) protected works which then could jeopardise
human artists ability to earn, put their livelihoods at risk and infringe their self-expression.82
The CAB allows for fair use for purposes including parody, homage, and pastiche.83 A pastiche
is an imitation of the style of another work or a collage of different works. AI-generated works
are all imitations of styles of existing works or amalgamations thereof and thus pastiches. Fair
use of works for the purpose of pastiche is a general exception from copyright protection. To
determine whether AI-generated works of pastiche amount to fair use many factors would be
considered, such as the amount of work copied and the nature of the work.84 AI-generated
works will not amount to infringements of the copyrighted works that the AI learns from and
imitates where it is determined to be fair use.

Scholars in South Africa have stated that the CAB should be redrafted to include explicitly that
AI-generated works cannot be protected under copyright. It should explicitly include that
human skill and effort (human authorship) are required for copyright protection. This would
protect human artists interests and continue to incentivise human creation. A further policy

77
Copyright Amendment Bill (“CAB”) (B13-2017).
78
Ibid at s2.2.
79
A Rens & H Hlomani ‘AI and the Copyright Amendment Bill – Research ICT Africa’ available at
https://pmg.org.za/committee-meeting/36546/ accessed on 20 April 2023.
80
Hanani Hlomani & Andrew Rens ‘AI and the Law Op-Ed - Artificial intelligence, copyright infringement and
protection – a legal quagmire?’ available at https://www.dailymaverick.co.za/article/2023-03-24-artificial-
intelligence-and-copyright-a-legal-quagmire/ accessed on 20 April 2023.
81
Ibid.
82
Alberts op cit note 18.
83
CAB op cit note 77 at s12A(a)(v).
84
Ibid at s12A(b).

15
recommendation includes a flexible provision enabling the use of copyrighted works within AI
research, including an opt-out mechanism, transparency requirements and a balancing test to
reduce harm to creators.85

IV. DOES COPYRIGHT LAW EXTEND TO AI-GENERATED

WORKS? – A COMPARATIVE LAW EXERCISE.
(a) Existing precedents around the world
In South Africa there is no precedent regarding whether AI-generated works are eligible for
copyright protection nor who would be considered the authors and owners of such works. I
must analyse other countries approaches and case law to guide me. By considering the practice
regarding copyright protection, authorship, and ownership of AI-generated works in other
countries, I can flag possible problem areas as well as solutions which may be useful within
the South African context. South African courts may consider foreign law when interpreting
the Bill of Rights. The Constitution allows the courts to use comparative law.86

(i) The US
The purpose of copyright in the United States is to incentivise progress in the sciences and arts
by granting creators exclusive rights over their works for a set time.87

a) Laws around copyrighting works

Under s102 of the Copyright Act of 197688 it is set out that copyright is granted to “original
works of authorship” which are in a fixed tangible medium.89 Originality within the US context
includes skill, labour, judgement as well as creativity.90 The US Code does not contain explicit
provisions regarding AI generated works and copyright.

85
A Rens, H Hlomani & S Msipa ‘Clarifying copyright to enable AI research in Africa’ available at
https://researchictafrica.net/publication/ai-and-intellectual-property-brief-1/ accessed on 29 April 2023.
86
The Constitution op cit note 32 at s39(1)(c).
87
The Constitution of the United States,1789 Article I, Section 8, Clause 8.
88
United States Code: Copyright Office, 17 U.S.C. §§ 201-216 (1958).
89
The Copyright Act of 1976 at s102.
90
D Oriakhogba ‘The scope and standard of originality and fixation in Nigerian and South African copyright
law’ (2018) African Journal of Intellectual property 2(2) 119-135, 123.

16
b) How Naruto v Slater may inform the AI copyright debate.
Naruto v Slater91 was a copyright case involving Naruto (a monkey) who used Slater’s (a
photographer) camera to take images of himself. These images were used by Slater in a book
which stated that the images were taken by Naruto without outside help.92 People for the Ethical
Treatment of Animals ("PETA") sued on behalf of Naruto for copyright infringement.93 The
United States Court of Appeals for the Ninth Circuit dismissed the case stating that Naruto as
an animal (a non-human) did not have legal personality or standing and cannot sue or pursue
claims for copyright. It was decided that animals (non-humans) may not be considered authors
of creative works, nor could they own copyright to the same.94 This view has been reinforced.95

Human authorship is required for copyright protection and autonomously AI-generated works
would not meet this requirement. The court’s ruling in Naruto resulted in the images being
released into the public domain as Slater also did not meet the authorship requirements.96

c) US approach to whether copyright applies to AI-generated works.

The Compendium of Best Practices by the U.S. Copyright Office (USCO) has stated that
human authorship is required for copyright protection.97 In USCO’s guide to registering
content containing AI-generated material they state that in the US Constitution and Copyright
Act the term “author” fundamentally excludes non-humans.98 AI is a machine, non-human, and
thus under the current precedents set in US copyright law AI-generated works will not meet
the authorship requirement and may not be granted copyright. USCO has been denying
copyright protection requests for AI-generated works.99

For example, Stephen Thaler’s “creativity machine” algorithm generated an image titled ‘A
Recent Entrance to Paradise’ which he sought to register for copyright protection. He was
denied as the work was produced by the AI without any creative intervention on his part. The

91
Naruto v. Slater, No. 16-15469 (9th Cir. 2018).
92
CB Ncube & DO Oriakhogba ‘Monkey Selfie and Authorship in Copyright Law: The Nigerian and South
African Perspectives’ (2018) PER / PELJ 2.
93
Ibid at 2.
94
A Guadamuz ‘The monkey selfie: copyright lessons for originality in photographs and internet jurisdiction’
(2016) Internet Policy Review, 5(1).
95
Naruto v. Slater supra note 91; People v. Frazier, 173 Cal. App. 4th 613 (2009); Urantia Found v. Kristen
Maaherra, 114 F.3d 955, 957-59 (9th Cir. 1997); Kelley v. Chicago Park Dist., 635 F.3d 290, 304 (7th Cir.
2011).
96
K Hristov ‘Artificial Intelligence and the Copyright Dilemma’ (2016) IDEA: The IP Law Review 448-449.
97
U.S. Copyright Office, Compendium of U.S. Copyright Office Practices § 101 available at
https://www.copyright.gov/comp3/docs/compendium.pdf accessed on 16 April 2023.
98
United States Copyright Office ‘Copyright Registration Guidance: Works Containing Material Generated by
Artificial Intelligence’ (2023) Federal Register, Vol. 88, No. 51 Rules and Regulations 37 CFR Part 202.
99
Congressional Research Service ‘Generative Artificial Intelligence and Copyright Law’ available at
https://crsreports.congress.gov/product/pdf/LSB/LSB10922 accessed on 11 April 2023.

17
AI is a non-human, so it did not meet the threshold for authorship. Thaler argued that non-
human generated works could be protected under copyright using the work-for-hire doctrine.
However, this doctrine applies only to employees or those with a work-for-hire agreement. It
is based on contracts and AI cannot enter legally binding contracts.100 The Review Board of
USCO decided the work could not be registered for copyright protection.101

Another application involved Kris Kashtanova’s graphic novel. Kashtanova used an AI to

produce the images (giving prompts) while she provided the storyline and layout of the AI-
generated images. USCO decided that the individual AI-generated images could not be
afforded copyright protection as they were not authored by a human. However, the graphic
novels storyline and arrangement could be afforded copyright protection as the text, the
storyline and the arrangement were human authored.102

USCO’s registration of AI-generated policy as of March 2023 is as follows. Their criteria

includes whether the work is mainly human authored with the tool-like assistance of
AI/computers. They further include whether the AI executed the traditionally human authored
elements of the work such as the literary or artistic expression, arrangement, and selection.
Where the traditional authorship elements were produced by AI the resultant works may not
receive copyright protection.103 An example would be where the AI receives only a prompt
from the human and creates a complex image. Where an AI has the ultimate creative control
there may be no copyright protection.104

However, where the human has taken AI-generated content and rearranged it in a creative
manner, the human-authored aspects will be afforded copyright protection to the extent that it
will not affect the original works copyright status. In essence, the Office will examine if the
human had enough creative control over the expression of the work and completed the
traditional authorship elements themselves.105

100
U.S. Copyright Office Review Board ‘Re: Second Request for Reconsideration for Refusal to Register ‘A
Recent Entrance to Paradise’ available at https://www.copyright.gov/rulings-filings/review-board/docs/a-
recent-entrance-to-paradise.pdf accessed on 16 April 2023.
101
Ibid.
102
Congressional Research Service op cit note 99 at 2.
103
United States Copyright Office op cit note 97.
104
Ibid.
105
Ibid.

18
Further, there are cases surrounding AI’s alleged copyright infringement such as the OpenAI
case106 and the Stability AI case.107 These pending cases address issues concerning the training
datasets containing copyright protected materials which the AI uses to generate “new” content.
AI has the tendency to replicate parts of copyrighted content without providing credit,
plagiarizing from datasets.108 AI-generated works which amount to copyright infringement
would not be protectable under copyright law. The ethical issues here are important but are
outside of the scope of this paper and so will not be discussed further.

(ii) The UK
a) Does copyright law recognise AI generated works?
In the UK, as with South Africa, copyright automatically vests when work is created by its
author in a fixed material form if the work is “original”. Computer software/programs are
included under literary, dramatic, and musical works in s3 of Copyright Designs and Patents
Act 1988 (CDPA).109 The originality threshold is that the work must be the authors own
intellectual idea/creation.110 Originality requires the creator have used sufficient skill, labour
or industry while creating the works (‘sweat of the brow’).111 Authorship requires a human
author, as the CDPA refers to the author as a ‘person’.112 Computer-generated literary,
dramatic, musical, or artistic works are deemed to have been authored by the person who made
the necessary arrangements for the computer to create the works.113 The South African
Copyright Act has adopted this same approach.

b) Who, if anyone, owns the copyright to AI generated works?

The author of AI-generated works is whoever made the arrangements necessary for the AI to
generate the works such as the user providing the prompt/instructions. The owner of the AI-
generated works would also usually be the human user of the AI.114 Where the AI created the
works independently of human intervention, arrangements, or creativity then there would be

106
Silverman v. OpenAI Inc., N.D. Cal., No. 3:23-cv-03416; P.M. et al v. Openai LP et al., No. 3:23-cv-03199.
107
Andersen et al v. Stability AI Ltd. et al No. 3:23-cv-00201-WHO.
108
Ibid; Silverman supra note 106; Kyle Wiggers ‘The current legal cases against generative AI are just the
beginning’ available at https://techcrunch.com/2023/01/27/the-current-legal-cases-against-generative-ai-are-
just-the-beginning/ accessed on 20 April 2023.
109
Copyright Designs and Patents Act 1988 s3(1)(b).
110
A Rahmatian ‘Originality in UK Copyright Law: The Old “Skill and Labour” Doctrine Under Pressure’
(2013) IIC 44, 4–34.1-3.
111
Oriakhogba op cit note 90 at 3-4.
112
Copyright Designs and Patents Act op cit note 109 at s9(1).
113
Ibid at s9(3).
114
Ibid.

19
no author. The AI developers are authors of the AI algorithm. However, where they are not
directly creatively involved in the subsequent AI-generated works, they would not be
considered authors of such works. Where there is no human skill or labour involved, the AI-
generated work would not meet the originality requirement and would not be protected by
copyright.115

Section 9(3) of the CDPA was applied in the case of Nova Productions v Mazooma Games and
Others wherein the court had to decide who owned the copyright over screenshots of a
computer game. They were computer-generated composite frames, where all creative decisions
were made by the developer. The player who took the screenshots did not make the necessary
arrangements for their creation, the developers of the game did and thus the developers owned
the copyright.116

Further, the court in Express Newspapers stated the difference between computer-generated
and computer-assisted/aided works as previously discussed under Payen v Bovic. This case set
out that where a human has used their own labour and skill as well as a computer to develop a
work, the computer is seen as a tool. While the human is seen as the author and thus the owner
of the copyright.117

Thus far the UK has not dealt with any specific cases in court regarding AI-generated works
and who owns the copyright therein.

c) Proposed AI Strategy
The UK’s National AI Strategy contains a ten-year plan which aims to achieve three key
actions.118 These actions are to invest in and plan for the long-term AI ecosystem needs to
continue their leadership as an “AI superpower”.119 They aim to support and encourage the
transition to an AI-enabled economy and to make sure that the UK governs AI technologies
correctly nationally and internationally.120 The UK government has committed to consult
through the IPO (Intellectual Property Office) on whether and to what extent AI-generated

115
Ibid at s2.
116
Nova Productions Ltd v Mazooma Games Ltd [2007] EWCA Civ 219.
117
Express Newspapers supra note 70.
118
HM Government ‘National AI Strategy’ available at
https://assets.publishing.service.gov.uk/government/uploads/system/uploads/attachment_data/file/1020402/Nati
onal_AI_Strategy_-_PDF_version.pdf accessed on 3 July 2023 1.
119
Ibid at 7.
120
Ibid at 7-9.

20
works should be protected under copyright law. They will also outline measures to ensure
easier use of copyright protected works for AI development.121

(iii) Germany
a) Copyright protected creative works.
Germany is part of the European Union (“EU”). The European Union's Court of Justice (CJEU)
has followed a human centred approach to copyright law. The CJEU determined the factors for
assessing whether a work should be protectable under copyright as being "free creative
choices", "the author's intellectual inventiveness", "personal touch of the creator" or
"personality of the creator".122

The German Copyright Act provides that a work may only be protected under copyright law
where it is original (the authors intellectual creation).123 As a human must have been
responsible for the intellectual creation, AI as well as corporations lack the ability to be authors
in copyright law in Germany.124 Further, the work must have been in a form which allows for
it to be objectively identifiable, even where the form is not permanent.125

b) How this translates to copyrightability of AI generated works

The programmers who train the AI and create the algorithms have a creative hand in making
the AI. However, they do not necessarily have the intellectual creative control over the works
that the AI generates. The programmer would only be considered the author where they had
sufficient creative input into the works appearance.126 Thus, under the existing German laws
the AI-generated works would need to be assessed separately to the original programming of
the AI.127

Preparatory acts are not protected under copyright law. The programmer selecting training data
inputs will not amount to them being declared an author of the AI-generated works. However,

121
Ibid at 42.
122
Eva-Maria Painer v Standard Verlags GmbH C-145/10 2011 at §§ 87-92; Football Dataco and Others v.
Yahoo! UK Ltd and Others C-604/2010 2012 at §38; Shuruq Aref Alshammari ‘The Copyright Protection of AI-
Created works by the European Union Copyright Legislation’ (PhD thesis, Aljouf University, 2021) 11-12.
123
Gesetz über Urheberrecht und verwandte Schutzrechte (“UrhG”) (The German Copyright Act).
124
J Freialdenhoven, N Maamar, S Mroß, Prof. Dr. JB Nordemann ‘Do AI generated works qualify for
copyright? Summary of the German report to the AIPPI 2019 Study Question on copyright in artificially
generated works’ (2020) IntellectualProperty.
125
Levola Hengelo BV/Smilde Foods BV C-310/17 2018 at §36.
126
Court of Appeal (Oberlandesgericht) Karlsruhe, dec. of 14-04-2010, docket-no. 6 U 46/09, GRUR-RR 2010,
234.
127
Freialdenhoven op cit note 124.

21
the data selected for the training can be protected under copyright as a database.128 Where a
person is involved in the selection and arrangement of AI-generated works as part of the
creative process then the resultant work would be copyrightable. The AI would be a ‘tool’
towards human creative achievements. Where the AI is responsible for all steps of the creative
process no copyright may vest in the work.129

German legal scholars have suggested that there should be incentives (new related rights) to
promote the creation of AI and the efforts of programmers. The legal scholars also warned that
people may lie about how much human creative control was involved in AI-generated works
to be granted copyright protection. Thus, measures should be arranged to counteract this.130

c) EU Approach and The AI Act

One of the highlighted issues within the EU concerns the training of AI systems using
copyrighted works. AI training using datasets is regulated in the Directive of Copyright in the
Digital Single Market (CDSM). Through two text and data mining (TDM) exceptions for
scientific purposes in Article 3 and commercial purposes in Article 4.131 For AI learning models
copyright infringement concerns the commercial TDM exception is applicable. Article 4
provides an exception for reproductions and extractions of works accessed lawfully for text
and data mining.132 The commercial TDM exception also provides for right holders to be able
to opt-out, so their work is not included within the data mining.133 This is already being used
within AI systems such as with Stable Diffusions AI tool company allowing creators to opt-
out of the datasets.134

The European Parliament is in the process of legislating an AI Act which seeks to implement
transparency requirements. These will require that AI-generated content be disclosed, to design
models to prevent them from generating illegal content. In addition, they will require for AI

128
UrhG op cit note 122 at s4, s87a.
129
Freialdenhoven op cit note 124.
130
Ibid.
131
JP Quintais ‘Generative AI, Copyright and the AI Act’ Institute for Information Law, Wolters Klewer (2023)
available at https://copyrightblog.kluweriplaw.com/2023/05/09/generative-ai-copyright-and-the-ai-act/ accessed
on 20 April 2023.
132
Ibid.
133
Ibid.
134
Ibid.

22
and AI training tool companies to publish summaries of all copyrighted data to be used for
training.135

(iv) Australia
a) Copyright Law in Australia
Copyright in Australia is regulated by the Copyright Act 1968. It provides an avenue for
creators to gain economic benefits from their work by protecting their creations as their own
personal property.136 It is aimed at rewarding and incentivising human creativity.137 Copyright
protection vests automatically in works which are in a material form and are original whether
the work has been published or not. Originality requires human authorship and creativity.138

b) Acohs Pty Ltd v Ucorp Pty Ltd

In the case of Acohs Pty Ltd v Ucorp Pty Ltd there was a claim by Acohs that the Material
Safety Data Sheets (MSDS) that it provided to companies were protected under copyright.139
Acohs maintained the necessary information for the creation of the MSDSs in a database. The
MSDSs were generated by a computer software system based on data provided by the
employees. Sometimes the data fed to the system was transcribed from existing MSDSs
produced by the software system. A customer would request an MSDS, and the employee
would send the request to the software system. The software system would then sort through
the database and compile source codes to create the correct MSDS.140

The Federal Court held that where the Acohs employees wrote the MSDSs (selected and
compiled the material) that the MSDSs were original literary works and could be protected

135
European Parliament News ‘EU AI Act: first regulation on artificial intelligence’ available at
https://www.europarl.europa.eu/news/en/headlines/society/20230601STO93804/eu-ai-act-first-regulation-on-
artificial-
intelligence#:~:text=Parliament's%20priority%20is%20to%20make,automation%2C%20to%20prevent%20har
mful%20outcomes accessed on 20 April 2023.
136
Copyright Act 1968 s196(1).
137
Australian Copyright Council ‘Submission to the UN Special Rapporteur on the International Covenant on
Economic, Social and Cultural Rights on the Impact of Intellectual Property Regimes on the Enjoyment of
Rights to Science and Culture’ available at
https://www.ohchr.org/sites/default/files/Documents/Issues/CulturalRights/ConsultationIntelectualproperty/Aust
ralian_copyrights_council.pdf accessed on 20 April 2023.
138
A Christie ‘Simplifying Australian Copyright Law - the Why and the How’ (2000) Australian Intellectual
Property Journal.
139
Acohs Pty Ltd v Ucorp Pty Ltd [2012] FCAFC 16.
140
Ibid.

23
under copyright. However, where the employees transcribed past MSDSs and did not make
any original contributions then there was no originality and no copyright protection.141

In addition, on appeal the Full Federal Court agreed. The Court stated that the software
programmers were not involved in the creation of the MSDSs. The programmers did not author
the MSDSs. The HTML source codes which underly the MSDSs were the creation of the
program itself in a routine using data from the database. The routine was coded by the
programmers, but this does not equate to the programmers authoring the source codes.142

Therefore, it was shown that where a computer program is responsible for the work without
human creativity involved there cannot be said to have been originality in the work. Thus, the
computer-generated work cannot be protected under Australian Copyright Law which requires
human authorship.143

(v) China
a) Copyright Law in China
The Copyright Law of People's Republic of China was enacted to protect the rights of authors
to encourage further creation, to better their socialist society through promoting the growth of
their cultures and sciences.144 Chinese citizens, unincorporated organisations and legal entities
may own the copyright over their works.145 Works are defined as being intellectual
achievements in fixed certain forms such as writing, art and computer software which have
originality.146 Copyright holders may be authors (natural persons) or otherwise legal persons
or unincorporated organisations where the works were made on behalf of them or where they
bear the responsibility or preside over the works.147

b) Shenzhen Tencent v. Shanghai Yingxun

In the case of Shenzhen Tencent v Shanghai Yingxun, an AI writing aid called “Dreamwriter”
was licenced to Shenzhen Tencent. Shenzhen Tencent posted an article by Dreamwriter to
Tencent Securities website with a notice to explain that it was written by their AI, Dreamwriter.

141
A Fitzgerald & N Dwyer ‘Copyright in databases in Australia’ available at
https://eprints.qut.edu.au/50425/4/50425.pdf accessed on 10 April 2023 3-4.
142
Acohs supra note 139.
143
Fitzgerald op cit note 141.
144
Copyright Law of People's Republic of China 1990 (as amended 2002) (中华人民共和国著作权法) Art 1.
145
Ibid at art 2.
146
Ibid at art 3.
147
Ibid at art 11.

24
Shanghai Yingxun reprinted this article without permission. Shenzhen Tencent then sued for
copyright infringement and unfair competition.148

There were two relevant issues, namely whether AI-generated works may be protected under
copyright and if so, who would own this copyright? The Court noted that the Dreamwriter
article constituted a written work under copyright law. The reasoning was that the creative team
at Tencent had a lot of control. They selected and arranged data inputs, provided the style
choice, template and the trigger conditions. These intellectual creative decisions were
expressed in the final article.149

Furthermore, the final presentation was determined by the creative team and Tencent. Thus,
they were the authors of the article and Dreamwriter was a tool to assist them. The work had
sufficient originality. As copyright law allows for protection of AI-assisted works and not
autonomously AI-generated works, this article was deemed to be protectable. Autonomously
AI-generated works have not yet been discussed in China’s courts, as existing AI is usually not
free from human intervention. They are mostly prompt based, and the data provided to the AI
is given by humans.150

The Copyright Law stipulates in Article 11 that a legal entity or unincorporated organisation
shall be deemed the author of any works where the works were created in accordance with their
will, under their sponsorship and where they have responsibility.151 The work was created by
Tencent teams (the authors) working under Shenzhen Tencent’s will, who presided over the
work and bore responsibility for the article. The owner of the copyright was deemed to be
Shenzhen Tencent.152

c) Gao Yang v. Youku

In Gao Yang v Youku case the Plaintiff used a camera containing software programmed to take
certain shots for him once it was on the hot air balloon. Although while airborne there was no
human intervention, it was not acting without human creative intervention. The Plaintiff had
previously selected factors such as the sensitivity and camera angles. Despite the camera

148
Shenzhen Tencent Computer System Co., Ltd. (“Shenzhen Tencent”) v Shanghai Yingxun Technology Co.,
Ltd. (“Shanghai Yingxun”), Nanshan District People's Court, Shenzhen, Guangdong Province on December 24,
2019.
149
Ibid.
150
Zhou Bo ‘Artificial Intelligence and Copyright Protection -- Judicial Practice in Chinese Courts’ available at
https://www.wipo.int/export/sites/www/about-
ip/en/artificial_intelligence/conversation_ip_ai/pdf/ms_china_1_en.pdf accessed on 30 April 2023.
151
Copyright Law op cit note 144 art 11.
152
Shenzhen Tencent supra note 148.

25
software picking out screenshots from the video footage, the human Plaintiff had enough
control and human intelligent creative intervention to satisfy the originality requirement. The
Plaintiff was considered the author of the work and awarded copyright ownership.153

According to a World Intellectual Property Organization (WIPO) report, AI is meant to

embody the idea that machines can create things and perform tasks which previously required
human intervention, without or with limited human intervention.154 Zhou Bo, a senior judge in
China, states that only when AI is evolved enough to require no human intervention, the law
will need to be adjusted. As far as current legal practice in China is concerned, AI-generated
works are the result of humans creative and intellectual activities which are performed by the
AI as the human’s assistant for AI outcomes. Thus, the current copyright laws being human
author centred is not an issue and the law can continue to apply without needing any changes.155

V. SHOULD AI-GENERATED WORKS BE PROTECTABLE

UNDER COPYRIGHT LAW IN SOUTH AFRICA?
It can be seen from the analysis of the different countries approaches to AI-generated works
under copyright law that copyright is generally accepted as being human-centred. All five
require human authorship for copyright protection to be granted and so have determined that
autonomous AI-generated works may not be protected under copyright law. It is also clear that
where an AI-generated work infringes on the copyright of works within the training dataset it
would not be protected under copyright law if it is not sufficiently original. AI was regarded
as unable to be a copyright owner as it has no legal personality. AI is neither a natural nor
juristic person. It is also clear that the author of a work is not necessarily also the copyright
owner especially where the author is an employee of a juristic person.

There are different approaches as to how much human intervention, creativity or control is
required in AI-generated works before the authorship and originality requirements are met. In
the US, the traditional authorship elements must have been completed by the human involved.
These elements include artistic or literary expression, compositions, arrangements, and

153
Gao Yang v Youku Beijing Intellectual Property Court (2017) Jing 73 Min Zhong No. 797 Civil Judgment.
April 2, 2020.
154
WIPO/IP/AI/2/GE/20/1 REV available at
https://www.wipo.int/edocs/mdocs/mdocs/en/wipo_ip_ai_2_ge_20/wipo_ip_ai_2_ge_20_1_rev.pdf accessed on
30 April 2023 para 11.
155
Zhou Bo op cit note 150.

26
selection.156 Where these elements are completed by AI, there is no copyright protection over
the works produced. In using this, they determined that a simple prompt is not enough. The
work created by AI is more complex and so the AI completed the authorship elements rather
than the human prompter.157 In Australia, where the human prompt is a transcription of work
already completed by the AI there is not enough human creativity involved for originality and
human authorship and therefore there is no copyright protection.158

However, in the UK and China a simple prompt/instruction by a human to the AI is enough for
human authorship. The prompter or their employer would be granted the copyright ownership
over the resultant works.159 China’s legal approach is based on the idea that there are currently
no AI-generated works which are truly autonomous and that there is usually human
involvement (human author).160 Thus, the humans involved or the company they work for
would be given ownership over the copyright protected works.161 In all five countries (even
where the AI-generated work itself is declared unprotectable) compositions, selections or
arrangements of AI-generated works are protectable under copyright.

In South Africa, copyright is also human-centred and human authorship is required.162 The
difference between purely computer-generated versus computer-aided/assisted works has been
acknowledged by South African courts.163 This suggests that AI-assisted works would be
copyrightable as the AI is used as a tool by the human creator. Whereas autonomously AI-
generated works would not be copyrightable.164

However, there are no set factors or clear amount of human intervention needed to determine
whether copyright ownership may be granted to the human or juristic person involved. Should
South Africa follow the US and Australian approach where certain prompts/instructions are
not sufficient and thus the AI-generated works are not protectable under copyright? Otherwise,
should South Africa follow the approach taken by the UK and China where there is a lower
threshold for the level of human intervention needed for copyright protection?

156
United States Copyright Office op cit note 97.
157
Ibid.
158
Acohs supra note 139.
159
Express Newspapers supra note 70; Zhou Bo op cit note 150.
160
Zhou Bo op cit note 150.
161
Shenzhen Tencent supra note 148.
162
Copyright Act op cit note 39 at s3(1).
163
Payen Components supra note 67.
164
Ibid.

27
South Africa mirrors the UK’s copyright laws in so far as the added provision within the
definition of ‘author’.165 This set out that the person who made the necessary arrangements for
the AI-generated works is the author.166 However, the meaning and scope of ‘necessary
arrangements' has yet to be interrogated within South Africa. I posit that a multi-factor test
should be produced to decide whether there has been sufficient human involvement such as the
US using the authorship elements. A useful factor to consider would include the extent of the
human’s contribution, skill, judgement, or labour involved in the creation of the work.

Otherwise, a test similar to the causation test within the law of delict could be useful in this
regard to determine who the author is. It can be asked but-for the human’s involvement would
this work have been generated by the AI. Thereafter, one could ask whether this link is
sufficient for copyright protection to be granted.167 If two people put the same prompt into the
same AI they will get different AI-generated works. This suggests that the humans are not the
proximate cause of the AI-generated works as the AI itself makes the bulk of the creative
decisions. If this is the case the link between the human intervention and the AI-generated work
is not sufficient for copyright protection to be granted.

The justifications for copyright law include that copyright provides exclusive legal rights over
creations from one’s own labour168 and creativity as a reward to incentivize creation thus
rewarding social utility.169 Copyright also protects creator’s moral rights, dignity, and self-
expression.170 It creates a limited monopoly to incentivise creation and avoid a market
failure.171 As previously discussed, there is no need to provide copyright protection as an
incentive for AI-generated works as AI will create regardless of moral and economic incentives
if it is prompted. AI is a machine and does not have moral rights nor does it require financial
incentives. AI will continue to be developed without copyright protection over the works it
generates being granted to the AI creators. They have enough of an incentive as they get paid
to develop the AI while the company that they work for is granted ownership over the copyright
of the AI algorithm itself.

165
Copyright Designs and Patents Act op cit note 109 at s9(3).
166
Copyright Act op cit note 39 at s1.
167
A Fagan ‘Causation in the Constitutional Court: Lee v Minister of Correctional Services’ (2013)
Constitutional Court Review 108, 123
168
Aplin & Davis op cit note 24 at 4-7.
169
Longan op cit note 21.
170
Du Bois op cit note 28 at 27-28.
171
Ibid.

28
Further, copyright must result in the greatest net happiness by balancing the least amount of
copyright protection necessary to promote creation without stifling the publics use and
enjoyment of IP. I argue that the best way to achieve the purposes set out within the
justifications is to only grant copyright protection to AI-generated works where there is
sufficient human skill, labour, and judgement involved.

The term ‘necessary arrangements’ within the South African Copyright Act should be
determined to be a higher threshold than providing a short simple prompt. If small unoriginal
human interventions within AI-generated works allow those humans to gain copyright
protection over the works, this would unnecessarily stifle public use and enjoyment over these
works. Copyright should not be unnecessarily extended to works which were not originally
intended to be protectable as this can cause a diminishment of the commons.172 The CAB could
be expanded to include regulations for AI like those within the CDSM in the EU.173 This would
help to reduce instances of copyright infringement on the part of AI through transparency
requirements and added opt-out mechanisms for copyright protected works within training
datasets.174

In South African copyright law, time and effort is not enough for originality. Skill and labour
must be involved. The skill, judgement or labour must have been substantial so not all prompts
should amount to skill and labour.175 If all prompts were accepted as such then every AI-
generated work would become the IP of the user regardless of whether there was sufficient
skill or labour. The effect of this would be that the number of raw materials within the commons
would be greatly diminished and human creation would be negatively affected.176

Thus, the suggested multi-factor test would aid in protecting the spirit and purpose of copyright
law in South Africa. AI-generated works should not be protectable where the AI was
responsible for the majority of the skill, labour and judgement involved in the creation of the
work. This would not be in line with the authorship and originality requirements set forth.

172
van der Walt & du Bois op cit note 52 at 44-46.
173
Quintais op cit note 131.
174
Ibid.
175
Haupt t/a Softcopy supra note 57 para 35.
176
van der Walt & du Bois op cit note 52 at 44-46.

29
VI. CONCLUSION
The main aim in this paper was to examine whether AI-generated works should be protected
through copyright law. This issue was explored using justificatory theories of the purpose of
copyright law. Some theories posit that copyright law exists to incentivise humans to create
works. Copyright does so by giving people exclusive rights over their works, protecting their
moral rights, and allowing them to monetize their works to earn a living. I argued that AI has
no need for these incentives nor any other incentives to promote creation. AI is not human and
has no need for money nor does it have any moral rights. AI creators are already incentivized
by their employers to continue creating AI itself. Thus, neither AI creators nor AI require
further incentives such as copyright over the AI-generated works.

I analysed the approaches of different countries to the question of whether AI-generated works
should be protectable under copyright and if so, who owns the copyright and who is the author.
In most of the countries the courts and the legislatures maintained that copyright law is a
specific protection afforded to works completed through the skill, labour, and
intellect/creativity of humans. Thus, there can be no copyright protection afforded to AI-
generated content created without any human intervention/authorship. This is in line with the
purpose of copyright being to incentivise human creation. AI as a machine cannot be said to
require copyright protection over works in which no human skill, labour or judgement was
involved as this does not meet the originality requirement.

However, China posits that truly autonomous AI-generated content without any human
intervention within the creation process does not exist. Thus, they believe there is no reason to
update copyright laws to handle questions of AI-generated content and copyright protection
yet. The UK and China draw a distinction between solely AI-generated content and AI-assisted
content. AI-assisted content is copyrightable as there is human authorship.

South Africa’s current legal approach seems to be aligned with the UK’s approach to AI-
generated content. However, I believe that there should be a multi-factor test similar to in the
US to determine whether the human intervention was sufficient to meet the authorship and
originality requirements under copyright law. It is unnecessary and against the purpose of
copyright law to protect works where the AI itself completed majority of the creative elements

30
while the human provided a broad simple prompt. It would be preferable for the public for such
works to become a part of the commons to encourage further human creation.

There is an opportunity within the CAB for South Africa to set out when AI-generated content
may be protected under copyright. It should exclude copyright protection for autonomously
AI-generated content and legislate clearly that copyright protection should only vest in works
where there has been sufficient human skill, judgement, and labour.

In essence, works created by AI without sufficient human input to meet the human authorship
and originality requirements should not be protectable under copyright. AI-generated works
should only be protectable where there was sufficient human judgement, skill, or labour
involved. Where this protection is granted the owner of such copyright would be the human
who used the AI as a tool within the creative process or their employer. AI itself cannot be
considered an author within copyright law as this would not align with the purpose of
copyright.

31
BIBLIOGRAPHY
LEGISLATION

Constitution of the Republic of South Africa, 1996.

The Constitution of the United States, 1789 Article I, Section 8, Clause 8.

The Copyright Act of 1976.

Copyright Act 98 of 1978.

Copyright Act 1968.

Copyright Amendment Bill B13B-2017.

Copyright Designs and Patents Act 1988.

Copyright Law of People's Republic of China 1990 (as amended 2002) (中华人民共和国著
作权法).

Gesetz über Urheberrecht und verwandte Schutzrechte (“UrhG”) (The German Copyright Act).

CASES

Acohs Pty Ltd v Ucorp Pty Ltd [2012] FCAFC 16.

Andersen et al v. Stability AI Ltd. et al No. 3:23-cv-00201-WHO.

Eva-Maria Painer v Standard Verlags C-145/10 GmbH 1 December 2011.

Express Newspapers PLC v Liverpool Daily Post & Echo PLC and others [1985] 1 WLR 1089
(Ch): [1985] FSR 306.

Football Dataco and Others v. Yahoo! UK Ltd and Others C-604/2010 1 March 2012.

Gao Yang v Youku Beijing Intellectual Property Court (2017) Jing 73 Min Zhong No. 797
Civil Judgment. April 2, 2020.

Haupt t/a Softcopy v Brewers Marketing Intelligence (Pty) Ltd 2006 4 SA 458 (SCA).

Kalamazoo Division (Pty) Ltd v Gay 1978 (2) SA 184 (C) p 192.

Kelley v. Chicago Park Dist., 635 F.3d 290, 304 (7th Cir. 2011).

32
King v SA Weather Service (716/2007) [2008] ZASCA 143.

Klep Valves (Pty) Ltd v Saunders Valve Co Ltd [1987] (2) SA 1 (A).

Levola Hengelo BV/Smilde Foods BV Cf. CJEU, dec. of 13-11-2018, docket-no. C-310/17.

Moneyweb (Pty) Ltd v Media 24 Ltd & another [2016] 3 All SA 193 (GJ); 2016 (4) SA 591
(GJ).

Naruto v. Slater, 2016 U.S. Dist. Lexis 11041 (N. D. Cal. Jan. 23, 2016).

Nova Productions Ltd v Mazooma Games Ltd [2007] EWCA Civ 219.

Payen Components SA Ltd v Bovic Gaskets CC 1995 (4) SA 441 (A).

People v. Frazier, 173 Cal. App. 4th 613 (2009).

P.M. et al v. Openai LP et al., No. 3:23-cv-03199.

Shenzhen Tencent Computer System Co., Ltd. (“Shenzhen Tencent”) v Shanghai Yingxun
Technology Co., Ltd. (“Shanghai Yingxun”), Nanshan District People's Court, Shenzhen,
Guangdong Province on December 24, 2019.

Silverman v. OpenAI Inc., N.D. Cal., No. 3:23-cv-03416.

Urantia Found v. Kristen Maaherra, 114 F.3d 955, 957-59 (9th Cir. 1997).

JOURNAL ARTICLES

Allen G ‘Understanding AI Technology’ Joint Artificial Intelligence Center 2020.

Alberts, I., Mercolli, L., Pyka, T. et al. Large language models (LLM) and ChatGPT: what will
the impact on nuclear medicine be?. Eur J Nucl Med Mol Imaging 50, 1549–1552 (2023).
https://doi.org/10.1007/s00259-023-06172-w.

Alberts W “Copyright in ideas: A lesson from the ivory tower” Juta’s Business Law 2008 Vol
16 part 2.

Alshammari SA ‘The Copyright Protection of AI-Created works by the European Union

Copyright Legislation’ (2021) Aljouf University.

Bhardwaj A, Khanna P, Kumar S, Pragya, ‘Generative Model for NLP Applications based on
Component Extraction’ (2020) Procedia Computer Science.

33
Bonadio E and McDonagh L ‘Artificial Intelligence as Producer and Consumer of Copyright
Works: Evaluating the Consequences of Algorithmic Creativity’ (2020) Intellectual Property
Quarterly.

Bosch J, Olsson HH, and Crnkovic I ‘Engineering AI Systems: A Research Agenda’ Artificial
Intelligence Paradigms for Smart Cyber-Physical Systems 2021 DOI: 10.4018/978-1-7998-
5101-1.ch001.

Boucher P ‘Artificial intelligence: How does it work, why does it matter, and what can we do
about it?’ (2020) EPRS | European Parliamentary Research Service.

Cetinic E and She J. 2022. Understanding and Creating Art with AI: Review and Outlook.
ACM Trans. Multimedia Comput. Commun. Appl. 18, 2, Article 66 (February 2022).
https://doi.org/10.1145/3475799.

Christie A, ‘Simplifying Australian Copyright Law - the Why and the How’ Australian
Intellectual Property Journal - Volume 11 2000.

Du Bois M "Justificatory Theories for Intellectual Property Viewed through the Constitutional
Prism" PER / PELJ 2018(21) - DOI http://dx.doi.org/10.17159/1727- 3781/2018/v21i0a2004
p 19-21.

Fagan A ‘Causation in the Constitutional Court: Lee v Minister of Correctional Services’

(2013) Constitutional Court Review.

Freialdenhoven J, Maamar N, Mroß S, Dr. Nordemann JB ‘Do AI generated works qualify for
copyright?’ IntellectualProperty Summary of the German report to the AIPPI 2019 Study
Question on copyright in artificially generated works 2020.

Ginsburg JC ‘People Not Machines: Authorship and What It Means in the Berne Convention’
(2018) IIC 49.

Guadamuz A ‘The monkey selfie: copyright lessons for originality in photographs and internet
jurisdiction’ (2016) Internet Policy Review.

Hassani H, Silva ES, Unger S, TajMazinani M, Mac Feely S. ‘Artificial Intelligence (AI) or
Intelligence Augmentation (IA): What Is the Future?’ AI (2020) 1(2).

Hristov, K, Artificial Intelligence and the Copyright Dilemma (September 1, 2016). IDEA: The
IP Law Review, Vol. 57, No. 3, 2017, Available at SSRN: https://ssrn.com/abstract=2976428.

34
Hoffer J.G., Ofner A.B., Rohrhofer F.M. et al. ‘Theory-inspired machine learning—towards a
synergy between knowledge and data’ (2022) Weld World 66.

Huston J, "Artificial Intelligence as a Content Creator in the Publishing Industry" (2022). Book
Publishing Final Research Paper. 65.
https://pdxscholar.library.pdx.edu/eng_bookpubpaper/65.

Longan M (2022) ‘A System Out of Balance: A Critical Analysis of Philosophical

Justifications for Copyright Law through the Lenz of Users’ Rights’. Michigan Journal of Law
Reform, 56. ISSN 0363-602X (In Press).

Matthews B ‘The Evolution of Copyright’ Political Science Quarterly, Vol. 5, No. 4

(Dec.,1890), pp. 583-602 Oxford University Press https://www.jstor.org/stable/2139530.

Ncube CB and Oriakhogba DO ‘Monkey Selfie and Authorship in Copyright Law: The
Nigerian and South African Perspectives’ (2018) PER / PELJ.

Oriakhogba, D ‘The scope and standard of originality and fixation in Nigerian and South
African copyright law’ African Journal of Intellectual property (2018) 2(2).

Onyeagbako V, ‘Justifications for Copyright and Patents Protection’ (January 20, 2020).
Available at
SSRN: https://ssrn.com/abstract=3596193 or http://dx.doi.org/10.2139/ssrn.3596193.

Rens A and Hlomani H, ‘AI and the Copyright Amendment Bill – Research ICT Africa’
Copyright Amendment Bill & Performers Protection Amendment Bill: public hearings 2023.

Rens, A., Hlomani, H, & Msipa, S. (2023). Clarifying copyright to enable AI research in Africa.
Generative AI and Intellectual Property Brief no. 1. Research ICT Africa.
https://researchictafrica.net/publication/ai-and-intellectual-property-brief-1/

Sheikh, H., Prins, C., Schrijvers, E. (2023). Artificial Intelligence: Definition and Background.
In: Mission AI. Research for Policy. Springer, Cham. https://doi.org/10.1007/978-3-031-
21448-6_2

United States Copyright Office ‘Copyright Registration Guidance: Works Containing Material
Generated by Artificial Intelligence’ 2023 16190 Federal Register, Vol. 88, No. 51 Rules and
Regulations 37 CFR Part 202.

35
van der Walt, AJ & du Bois, M ‘The importance of the commons in the context of intellectual
property’ (2013) 1 Stell LR.

Yang, D., Zhou, Y., Zhang, Z., Li, T. J-J., & LC, R. (2022). AI as an Active Writer: Interaction
strategies with generated text in human-AI collaborative fiction writing. In A. Smith-Renner,
& O. Amir (Eds.), Joint Proceedings of the IUI 2022 Workshops: APEx-UI, HAI-GEN,
HEALTHI, HUMANIZE, TExSS, SOCIALIZE (pp. 56-65). (CEUR Workshop Proceedings;
Vol. 3124). CEUR-WS Team. http://ceur-ws.org/Vol-3124/.

Yu, KP ‘The Objectives and Principles of the TRIPS Agreement’, 46 Hous. L. Rev. 979 (2009).
Available at: https://scholarship.law.tamu.edu/facscholar/457.

Zhuk A ‘Navigating the legal landscape of AI copyright: a comparative analysis of EU, US,
and Chinese approaches’ (2023) AI Ethics.

BOOKS

Aplin T & Davis J ‘Copyright I: history, justifications, sources of law, and subsistence’ in
Intellectual Property (2022).

Dean T; Allen, J.; Aloimonos, Y. Artificial Intelligence: Theory and Practice;

Benjamin/Cummings: New York, NY, USA, (1995).

Elkin-Koren, N ‘Copyright in a Digital Ecosystem: A User-Rights Approach’ in Okediji R (ed)

Copyright in an age of Limitations and Exceptions (2015).

Spence M, 'The Justification of Intellectual Property' in Intellectual Property, Clarendon Law

Series (2007).

INTERNET SOURCES (OTHER)

Australian Copyright Council ‘Submission to the UN Special Rapporteur on the International

Covenant on Economic, Social and Cultural Rights on the Impact of Intellectual Property
Regimes on the Enjoyment of Rights to Science and Culture’ available at
https://www.ohchr.org/sites/default/files/Documents/Issues/CulturalRights/ConsultationIntele
ctualproperty/Australian_copyrights_council.pdf accessed on 20 April 2023.

36
Anne Fitzgerald & Natasha Dwyer ‘Copyright in databases in Australia’ available at
https://eprints.qut.edu.au/50425/4/50425.pdf accessed on 10 April 2023.

Banoula M ‘How to Become an Artificial Intelligence (AI) Engineer in 2023?’ available at

https://www.simplilearn.com/tutorials/artificial-intelligence-tutorial/how-to-become-an-ai-
engineer, accessed on 23 April 2023.

Bo Z ‘Artificial Intelligence and Copyright Protection -- Judicial Practice in Chinese Courts’

IPR Division of the Supreme People’s Court of China available at
https://www.wipo.int/export/sites/www/about-
ip/en/artificial_intelligence/conversation_ip_ai/pdf/ms_china_1_en.pdf accessed on 30 April
2023.

Congressional Research Service ‘Generative Artificial Intelligence and Copyright Law’

available at https://crsreports.congress.gov/product/pdf/LSB/LSB10922 accessed on 11 April
2023.

Davenport TH & Mittal N ‘How Generative AI Is Changing Creative Work’ Harvard Business
Review 2022 available at https://hbr.org/2022/11/how-generative-ai-is-changing-creative-
work, accessed on 10 July 2023.

Duggan W ‘Artificial Intelligence Stocks: The 10 Best AI Companies’ available at

https://money.usnews.com/investing/stock-market-news/slideshows/artificial-intelligence-
stocks-the-10-best-ai-companies accessed on 1 July 2023.

European Commission’s High-Level Expert Group on Artificial Intelligence ‘A definition of

AI: Main capabilities and scientific disciplines’ available at
https://ec.europa.eu/futurium/en/system/files/ged/ai_hleg_definition_of_ai_18_december_1.p
df accessed on 20 July 2023.

European Parliament News ‘EU AI Act: first regulation on artificial intelligence’ available at
https://www.europarl.europa.eu/news/en/headlines/society/20230601STO93804/eu-ai-act-
first-regulation-on-artificial-
intelligence#:~:text=Parliament's%20priority%20is%20to%20make,automation%2C%20to%
20prevent%20harmful%20outcomes accessed on 20 April 2023.

37
Hlomani H & Rens A ‘AI and the Law Op-Ed - Artificial intelligence, copyright infringement
and protection – a legal quagmire?’ available at https://www.dailymaverick.co.za/article/2023-
03-24-artificial-intelligence-and-copyright-a-legal-quagmire/ accessed on 20 April 2023.

HM Government ‘National AI Strategy’ available at

https://assets.publishing.service.gov.uk/government/uploads/system/uploads/attachment_data
/file/1020402/National_AI_Strategy_-_PDF_version.pdf accessed on 3 April 2023.

Korf RE ‘Does Deep-Blue use AI?’ available at

https://www.google.com/url?sa=i&rct=j&q=&esrc=s&source=web&cd=&cad=rja&uact=8&
ved=0CAIQw7AJahcKEwjwyvfplMuBAxUAAAAAHQAAAAAQAg&url=https%3A%2F
%2Faaai-22.aaai.org%2FPapers%2FWorkshops%2F1997%2FWS-97-04%2FWS97-04-
001.pdf&psig=AOvVaw3z_78NpiJr8-
RCp2b9nI8s&ust=1695916848782607&opi=89978449 accessed on 3 April 2023.

Launchbury J ‘A DARPA Perspective on Artificial Intelligence’ available at

https://www.darpa.mil/attachments/AIFull.pdf accessed on 5 April 2023.

LexisNexis & Squire Sanders ‘International aspects of copyright’ available at

https://www.squirepattonboggs.com/~/media/files/insights/publications/2012/10/international
-aspects-of-copyright/files/international-aspects-of-copyright-
practice%20note/fileattachment/international-aspects-of-copyright-practice-note.pdf accessed
on 8 April 2023.

Quintais JP ‘Generative AI, Copyright and the AI Act’ Institute for Information Law, Wolters
Klewer (2023) available at https://copyrightblog.kluweriplaw.com/2023/05/09/generative-ai-
copyright-and-the-ai-act/ accessed on 20 April 2023.

Tiernan J ‘Is AI Really an Art and Design Killer?’ available at

https://www.vectornator.io/blog/will-ai-replace-artists/#what-dangers-might-ai-present-for-
designers-and-illustrators, accessed on 28 April 2023.

TRIPS: Agreement on Trade-Related Aspects of Intellectual Property Rights, Apr. 15, 1994,
Marrakesh Agreement Establishing the World Trade Organization, Annex 1C, 1869 U.N.T.S.
299, 33 I.L.M. 1197 available at https://www.wto.org/english/docs_e/legal_e/27-
trips_01_e.htm accessed on 10 April 2023.

38
U.S. Copyright Office, Compendium of U.S. Copyright Office Practices § 101 available at
https://www.copyright.gov/comp3/docs/compendium.pdf accessed on 16 April 2023.

U.S. Copyright Office Review Board ‘Re: Second Request for Reconsideration for Refusal to
Register A Recent Entrance to Paradise’ available at https://www.copyright.gov/rulings-
filings/review-board/docs/a-recent-entrance-to-paradise.pdf accessed on 16 April 2023.

Wiggers K ‘The current legal cases against generative AI are just the beginning’ available at
https://techcrunch.com/2023/01/27/the-current-legal-cases-against-generative-ai-are-just-the-
beginning/ accessed on 20 April 2023.

WIPO/IP/AI/2/GE/20/1 REV available at

https://www.wipo.int/edocs/mdocs/mdocs/en/wipo_ip_ai_2_ge_20/wipo_ip_ai_2_ge_20_1_r
ev.pdf accessed on 30 April 2023

WIPO ‘Summary of the Berne Convention for the Protection of Literary and Artistic Works
(1886)’ available at https://www.wipo.int/treaties/en/ip/berne/summary_berne.html#_ftn2
accessed on 25 April 2023.

WIPO ‘WIPO-Administered Treaties: Contracting Parties Berne Convention’ available at

https://www.wipo.int/wipolex/en/treaties/ShowResults?search_what=C&treaty_id=15
accessed 15 April 2023.

39