SOCIAL MEDIA CONTENT MODERATION AND LIMITATIONS ON
FREEDOM OF EXPRESSION: THE ROLE OF THE STATE

Tara Davis

Supervisors: Franziska Sucker an Mariya Badeva-Bright

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:30 October 2022

Abstract

Social media platforms are an important space for the exercise of freedom of expression. However,

their operation poses a concomitant risk to the right. The right is both protected by and

undermined by content moderation - a mechanism that enables platforms to determine the bounds

of permissible speech, and what expression to highlight or suppress. This has significant

implications for the right to freedom of expression but has been left largely to the private sector to

regulate. This raises important questions about the role of the state in protecting and enabling the

right to freedom of expression online. In this paper I discuss the need for content moderation, how

it operates and the ways in which it undermines the right to freedom of expression in South Africa.

I further explore the role of the state and analyse South Africa’s current regulatory response. In

so doing, I contend that South Africa’s approach does not adequately respect, protect and promote

the right to freedom of expression.

Contents

I. INTRODUCTION 4

II. THE UNREALISED PROMISE OF THE INTERNET AND WHY WE MODERATE 5

(a)The need for content moderation ................................................................................................... 5

(b) The rationale for moderation and how it is defined ...................................................................... 7

(c) The operation of content moderation ............................................................................................ 8

(d) Considerations for content moderation ...................................................................................... 12

III. THE RIGHT TO FREEDOM OF EXPRESSION IN SOUTH AFRICA 14

(a) Un-packing the right to freedom of expression in South Africa .................................................. 14

(b) The operation and limitations of the right to freedom of expression .......................................... 15

(c) Freedom of expression online ..................................................................................................... 17

IV. HOW CONTENT MODERATION UNDERMINES THE RIGHT TO FREEDOM OF

EXPRESSION IN SOUTH AFRICA 18

(a) The limitations extend beyond section 16(2) ............................................................................... 19

(b)The limitations are vague ............................................................................................................. 19

(c) Prior restraint ............................................................................................................................. 20

V. THE ROLE OF THE STATE 22

(a) Why involve the state? ................................................................................................................. 22

(b) Reconceptualising the right in South Africa ............................................................................... 24

VI. SOUTH AFRICA’S REGULATION OF ONLINE EXPRESSION 26

(a) Conditional limitation of liability ............................................................................................... 26

(b) The take-down notification process ............................................................................................ 27

(i) Incentivising removal .............................................................................................................. 28

(ii) Outsourcing rights-based determinations ............................................................................. 28

(c) The complaints mechanism ......................................................................................................... 29

(d) The obligation to classify ............................................................................................................ 30

VII. CONCLUSION 31

I. INTRODUCTION

In 2020, three black footballers received hundreds of racially abusive comments on their social

media profiles,1 many of which included an emoji of a monkey. Despite multiple reports of abuse,

a significant amount of the content remained online, despite its harmful, racist nature.2 In the same

year, activists from Nigeria who were protesting police brutality and the Nigerian government, had

their content blocked. Despite the public importance of disseminating the information, the content

was incorrectly flagged as false and removed from social media platforms.3

These examples demonstrate the need for content moderation online and its potential harm.

Abusive content – such as hate speech– may infringe on the rights of others and undermine their

safety online and offline. However, incorrect removal may amount to censorship and infringe the

right to freedom of expression and other associated rights. There has accordingly been significant

debate around the appropriate balance between protecting freedom of expression and mitigating

against harm online. This balance is mediated through content moderation – the process through

which platforms determine the bounds and operation of permissible content on their site. It is no

easy task and one which has been left largely to the private sector.

What these examples also allude to is the context within which content moderation occurs. A

context where moderation choices are dictated by power and influence, and sometimes political

interference. This context, coupled with the significant implications that content moderation has

on the right to freedom of expression raises questions about the appropriate role of the state in the

moderation process.

In this paper, I discuss the role of the state in limiting the implications of content moderation on

social media platforms on the right to freedom of expression in South Africa. In so doing, I focus

1 Victoria Elms & Kieran Devine ‘Euro 2020: Why is it so difficult to track down racist trolls and remove hateful
messages on social media?’ Sky News 21 July 2021 at 1, available at https://news.sky.com/story/euro-2020-why-is-it-
so-difficult-to-track-down-racist-trolls-and-remove-hateful-messages-on-social-media-12358392, accessed on 4
February 2022.
2 David Wicock ‘Instagram claims monkey emojis “DON’T breach race rules” as Twitter removes 1,000 racist posts
after appalling abuse of England stars’ Daily Mail 12 July 2021 at 1, available at
https://www.dailymail.co.uk/news/article-9780627/Racist-tweeters-knock-door-police-face-force-law.html, accessed
on 4 February 2022.
3 David Gilbert ‘Facebook and Instagram are Censoring Protests Against Police Violence in Nigeria’ Vice 21
October 2021 at 2, available at https://www.vice.com/en/article/jgqeyg/facebook-is-censoring-protests-against-
police-violence-in-nigeria, accessed on 4 February 2022.

https://news.sky.com/story/euro-2020-why-is-it-so-difficult-to-track-down-racist-trolls-and-remove-hateful-messages-on-social-media-12358392
https://news.sky.com/story/euro-2020-why-is-it-so-difficult-to-track-down-racist-trolls-and-remove-hateful-messages-on-social-media-12358392
https://www.dailymail.co.uk/news/article-9780627/Racist-tweeters-knock-door-police-face-force-law.html
https://www.vice.com/en/article/jgqeyg/facebook-is-censoring-protests-against-police-violence-in-nigeria
https://www.vice.com/en/article/jgqeyg/facebook-is-censoring-protests-against-police-violence-in-nigeria

on five things. First, content moderation – why we need it, what it is and how it works. Second,

the right to freedom of expression in South Africa – why it matters, how it may be lawfully limited,

and how it operates online. Third, the implications that content moderation has on the right to

freedom of expression – including how it limits the right. Fourth, the role of the state and finally,

South Africa’s current regulation – what it’s doing, and whether it's sufficient.

These focus areas culminate in a conclusion that the state’s regulatory measures inadequately

extinguish its duty to respect and protect the right to freedom of expression.

II. THE UNREALISED PROMISE OF THE INTERNET AND WHY WE MODERATE

(a)The need for content moderation

The internet has fundamentally changed the exercise of the right to freedom of expression. Its

revolutionary potential was attributed to its participatory nature as an interactive medium.4 Unlike

other communication technology, such as print and television, internet users are not confined to

the role of passive observer but can actively engage as creators and publishers of their own

content.5 The internet promised to elevate anyone’s content and remove the publisher as

gatekeeper.6 Further, engagement on the internet is inexpensive, does not require specialist

equipment, provides a large, global audience, and may provide a degree of anonymity.7 These

characteristics posed an opportunity for unprecedented, unconstrained freedom of expression,

making early users of the internet hopeful of its potential as an open, participatory and democratic

space.8

Years and experience, however, proved that early hope to be a fallacy. Instead, digital spaces were

overrun by harmful content and conduct including things such as spam, violence, hate speech,

abuse and harassment. According to Grimmelman, ‘that’s just how people act on the Internet.’9

4 Report of the Special Rapporteur on the Promotion and Protection of the Right to Freedom of Opinion and
Expression Doc A/HRC/17/27 16 May 2011 at 6.
5 Ibid.
6 Tarleton Gillespie Custodians of the internet: Platforms, content moderation, and the hidden decisions that shape
social media (2018) 16.
7 Fawzia Cassim ‘Regulating hate speech and freedom of expression on the internet: Promoting tolerance and
diversity’ (2015) 28(3) South African Journal of Criminal Justice 303 at 305.
8 Gillespie op cit note 6 at 16.
9 James Grimmelmann ‘The virtues of moderation’ (2015) 17 Yale Journal of Law and Technology 42 at 45.

Despite his rather glib remark, online abuse is a pervasive and significant problem that threatens

the safety, well-being, and rights of users.

Harmful content is prevalent online. In just 3 months Facebook acted on 13.5 million pieces of

content it considered hate speech,10 45.9 pieces of violent and graphic content and11 8.2 million

pieces identified as harassment.12 Some content is targeted at a specific user – where expression is

used to attack a user’s gender, sexual orientation or race.13 This is a form of harassment, which is

disproportionality experienced by women and racial minorities.14 Some content is disseminated

with the intent to harm a broader community, referred to as trolling.15 In these instances a user

intentionally posts antagonistic information to a forum or group to elicit a hostile response.16 In its

most extreme form, some expression entails an orchestrated and intentional attack by an entire

community aimed at harming society.17

Abusive content is not the only form of harmful content – the dissemination of false information

poses a challenge too. This is not a new phenomenon, but the internet has proliferated the

concern.18 Now, anyone can develop and share false information.19 Research suggests that people

prefer to access information shared on social media,20 and it spreads faster than correct news due

to its attraction and novelty.21

10 Facebook Community Standards Enforcement Report 2022 ‘Hate Speech’ available at
https://transparency.fb.com/data/community-standards-enforcement/hate-speech/facebook/, accessed on 1 October
2022.
11 Facebook Community Standards Enforcement Report 2022 ‘Violent and Graphic content’, available at
https://transparency.fb.com/data/community-standards-enforcement/graphic-violence/facebook/, accessed on 1
October 2022.
12 Facebook Community Standards Enforcement Report 2022 ‘Bullying and Harassment’, available
athttps://transparency.fb.com/data/community-standards-enforcement/bullying-and-harassment/facebook/, accessed
on 1 October 2022
13 Tarleton Gillespie ‘Platforms are not intermediaries’ (2018) 2.2 Georgetown Law Technology Review 198 at 207.
14 Ibid.
15 Grimmelmann op cit note 9 at 54.
16 Lior J Strahilevitz ‘Wealth without markets?’ (2007) 116 Yale Law Journal 1472 at 1493.
17 Grimmelman op cit note 9 at 54.
18 Maria D Molina, S. Shyam Sundar, Thai Le & Donwon Lee ‘“Fake News” is not simply false information: A
concept explication and taxonomy of online content’ (2021) 65(2) American Behavioural Scientist 180 at 180.
19 Alvaro Figueira & Luciana Oliveria ‘The current state of fake news: Challenges and opportunities’ (2017) 121
Procedia Computer Science 817 at 818.
20 Molina et al op cit note 18 at 183.
21 Alexandre Bovet & Hernan Makse ‘Influence of fake news in twitter during the 2016 US Presidential Election’
(2019) 10 Nature Communications 1 at 2.

https://transparency.fb.com/data/community-standards-enforcement/hate-speech/facebook/
https://transparency.fb.com/data/community-standards-enforcement/graphic-violence/facebook/
https://transparency.fb.com/data/community-standards-enforcement/bullying-and-harassment/facebook/
https://transparency.fb.com/data/community-standards-enforcement/bullying-and-harassment/facebook/

The consequences of these harms are not confined to the online realm and examples abound of

their effects in real life. GamerGate is an example of persistent and coordinated harassment

targeted at prominent women in the gaming industry. They were subjected to months of online

abuse that included threats of rape, physical harm and death.22 One woman fled her home after her

address was posted online and another went into hiding.23

Such harm is also not confined to the physical but can permeate and undermine important societal

functions, values, and rights. This is evidenced by the alleged impact of false information on the

outcome of elections.24 The orchestrated dissemination of false information at important moments

shapes public discourse which may influence the outcome of democratic decisions.

Such harms are further felt by a shrinking marketplace of ideas and the undermining of the right

to freedom of expression. The proliferation of harmful content and behaviour contributes to a toxic

culture on platforms.25 Consequently, many users may censor themselves, avoid particular content

or leave platforms.26 This has a chilling effect on freedom of expression and undermines the

potential of platforms to flourish as inclusive online communities.27

(b) The rationale for moderation and how it is defined

Content moderation is posited as a solution to these myriad harms.28 Prompting some to argue that

it is the difference between platforms that succeed and those that are abandoned.29

22 Kate Klonick ‘The new governors: the people, rules and processes governing online speech’ (2018) 131 Harvard
Law Review 1598 at 1628.
23 Emily VanDerWerff ‘#Gamergate: Here’s why everybody in the video game world is fighting’ Vox 2014,
available at https://www.vox.com/2014/9/6/6111065/gamergate-explained-everybody-fighting, accessed on 17
February 2021.
24 See for example reports on the 2016 United States of America election and the 2022 Kenyan election. Following
the 2016 USA election, it was reported that Russian operatives allegedly created 50, 258 Twitter accounts which
were linked to bots (fake accounts which were programmed to disseminate false information). See Sophie
Marineau‘Fact check US: what is the impact of russian interference in the US presidential election?’ The
Conversation 29 September 2020 available at https://theconversation.com/fact-check-us-what-is-the-impact-of-
russian-interference-in-the-us-presidential-election-146711, accessed on 26 November 2020.
25 Gillespie op cit note 13 at 207.
26 Ibid.
27 Gillespie op cit note 6 at 16.
28 Kyle Langvardt ‘Regulating online content moderation’ (2018) 106 Georgetown Law Journal 1353 at 1358.
29 Grimmelmann op cit note 9 at 45.

https://www.vox.com/2014/9/6/6111065/gamergate-explained-everybody-fighting
https://theconversation.com/fact-check-us-what-is-the-impact-of-russian-interference-in-the-us-presidential-election-146711
https://theconversation.com/fact-check-us-what-is-the-impact-of-russian-interference-in-the-us-presidential-election-146711

This is the rhetoric associated with moderation – it is conducted to prevent harm. Grimmelmann

accepts this motivation and includes it in his definition of moderation which he considers ‘the

governance mechanisms that structure participation in a community to facilitate cooperation and

prevent abuse.’30 However, several other factors influence the moderation policies of platforms

including regulatory compliance, profit maximisation and response to public outcry.31 Platforms

profit by collecting large amounts of data about users to sell targeted advertising.32 They

accordingly have an interest in increasing engagement because it creates more data. This motivates

platforms to promote content that is ‘emotionally charged, extreme and inflammatory’33 because

it spreads quickly and is viewed extensively. The prevention of harm is accordingly not the only

rationale for moderation, it may be intricately tied with business models that prioritise user

engagement, data surveillance and profit. 34 Because of this, I define moderation more neutrally.

I define it as the governance mechanisms that determine the bounds of permissible content and the

actions that may be taken in response to violations. In practical terms, moderation determines what

content to remove, what to highlight and what to suppress on a platform. Such decisions dictate

the marketplace of ideas, founding perceptions that platforms are ‘content gatekeepers35,

‘governors of online speech’36 and ‘custodians of the public sphere’.37 Such a process is at odds

with early ideals of the internet but is now widely conducted by platforms.

The operation of moderation entails the development of rules, the creation and training of systems

to enforce the rules and a process of iteration that relies on external reaction to update them.38 It

also involves a process to adjudicate on compliance with such rules, and responses to non-

compliance. Such a system is analogous to a governance or legal system39 and its operation has

30 Ibid at 47.
31 Barrie Sander ‘Freedom of expression in the age of online platforms: the promise and pitfalls of a human rights-
based approach to content moderation’ (2020) 43 Fordham International Law Journal 939 at 950–955.
32 Daphne Keller ‘Internet platforms: observations on speech, danger, and money’ (2018) Hoover Working Group on
National Security, Technology and Law, Aegis Series Paper No. 1807 at 4.
33 Sander op cit note 31 at 954.
34 Ibid.
35 Ibid at 941.
36 Klonick op cit note 22 at 1602.
37 Jack M Balkin ‘Free Speech is a Triangle’ (2018) 118 Columbia Law Review 2011 at 2022.
38 Ibid.
39 Klonick op cit note 22 at 1602.

aptly been characterised as a form of ‘private governance.’40 This conception is grounded on the

‘quasi-normative’, ‘quasi-executive’ and ‘quasi-judicial’ powers exerted by platforms when they

make the rules, enforce the rules, and determine compliance with the rules.41

(i) Quasi-normative powers

Platforms provide the rules for moderation in publicly accessible documents referred to as

‘community standards’ or ‘terms of use.’42 Users are typically required to accept the terms to

access the platform.43The rules dictate the type of content that is permitted and prohibited. Each

platform has different rules which are underpinned by the specific values of the platform’s

community.44 Although there are distinctions between platforms, there is significant overlap. For

example, Facebook, Twitter and TikTok, all prohibit the sexual exploitation of children and have

rules concerning hate speech and violence.

The rules generally apply to all users, regardless of their location as expressly provided by

Facebook45 and TikTok.46 However, some platforms treat content differently depending on the

jurisdiction. Twitter withholds content in certain countries if they receive a request from ‘an

authorised entity’ (such as a court order) or the content violates local laws.47 Access will only be

restricted in that territory. Access may also be age restricted.48

40 Balkin op cit note 37 at 2032.
41 Luca Belli, Pedro Augusto Francisco & Nicolo Zingales ‘Law of the land or law of the platform? Beware of the
privatisation of regulation and police’ in Luca Belli & Nicolo Zingales (eds) Platform regulations (2017) 41at 59.
42 Facebook’s content rules are referred to as Community Standards and TikTok refers to them as Community
Guidelines.
43 Report of the Special Rapporteur on the Promotion and Protection of the Right to Freedom of Opinion and
Expression Doc A/HRC/32/38 11 May 2011.
44 Klonick op cite note 22 at 1628.
45 Facebook’s Community Standards, available at https://transparency.fb.com/en-gb/policies/community-standards/,
accessed on 1 October 2022.
46 TikTok’s Community Guidelines, available at https://www.tiktok.com/community-guidelines?lang=en, accessed
on 1 October 2022.
47 Twitter Help Center ‘About country withheld content’, available at https://help.twitter.com/en/rules-and-
policies/tweet-withheld-by-country, accessed on 1 October 2022.
48 Facebook will only make certain content available for users over 18. See for example, Facebook’s Community
Guidelines on content concerning adult nudity and sexual activity.

https://transparency.fb.com/en-gb/policies/community-standards/
https://www.tiktok.com/community-guidelines?lang=en
https://help.twitter.com/en/rules-and-policies/tweet-withheld-by-country
https://help.twitter.com/en/rules-and-policies/tweet-withheld-by-country

Such rules also typically apply to all forms of expression–including videos, livestreams, pictures,

audio, links and text-based expression.49 Twitter enforces its rules against tweets, which are

publicly accessible, as well as direct messages, which are private.50

Such rules determine the bounds of permissible content by dictating what can and cannot be

expressed. This forms the first part of our definition. The second part – the actions that may be

taken against infringing content – includes the measures that platforms take against content that

violates their rules.

(ii) Quasi-executive powers

Moderation policies are enforced in several ways including removing the impugned content,

restricting access to it, temporarily or permanently suspending a user’s account, or temporarily or

permanently banning someone from the platform.51 The action taken by a platform is generally

tiered, depending on its severity. This is evident on Facebook where adult nudity and sexual

activity content, is either removed, placed behind a sensitivity-warning or restricted for individuals

under 18.52 On Facebook, TikTok and Twitter, suspension or banning is reserved for instances of

severe violations.53 In making such determinations they consider whether the user repeatedly

violated the rules, and TikTok considers ‘actions on other platforms and offline behaviour.’54

Platforms also respond to violating content by reducing its visibility. TikTok’s Community

Guidelines state ‘for some content, we may reduce discoverability, including by redirecting search

results, or making videos ineligible for recommendation.’55 This appears to be a highly

discretionary response. This is the case on Twitter too. Twitter provides very little information on

when or how decisions are made, and simply notes: ‘[l]imiting Tweet visibility depends on a

number of signals about the nature of the interaction and the type of content.’56 These provisions

49 TikTok’s Community Guidelines explicitly state that it applies to ‘everything on TikTok’.
50 Twitter op cit note 47.
51 Facebook op cit note 45.
52 Ibid.
53 TikTok for example, provides that they will ‘ban accounts and/or users that are involved in severe or repeated on-
platform violations.’ The banning of Ex-President Donald Trump from Twitter is a notorious example of this.
54 TikTok op cit note 46.
55 Ibid. Additional information is provided under ‘Ineligible for the For You-Feed’ which provides the types of
content which would be considered ineligible, these include: dangerous stunts and sports, overly sexualised content
and tobacco and alcohol products, amongst others.
56 Twitter op cit note 47.

allow platforms to choose what information to highlight, and what information to suppress;

ultimately determining what expression is valuable, and what shouldn’t be seen. Such decisions

underpin conceptions that they curate discourse.

The policies are enforced using a combination of human intervention and automation, and there

are different processes that may lead to action. 57 Content may be automatically flagged by

software and removed before it is ever published. 58 This is a form of prior restraint. It may be

published but then flagged as violating content by software or users, which is then reviewed.

(iii) Quasi-judicial powers

The wording of the rules affords platforms a high degree of discretion – to determine what

expression amounts to, whether it is a violation, and the appropriate response. Importantly, Twitter,

TikTok and Facebook all have a public interest override. This permits violating expression to

remain online if it is in the public interest. This is not an objective determination, and all three

platforms provide vague substantiation of its application. TikTok includes an exhaustive list of the

circumstances that warrant a deviation from removal for offending content, which includes satire,

counter-speech and content that ‘enables individual expression on topics of social importance.’59

It is not clear what this means. In determining the public interest, Facebook weighs up whether

there is a threat to safety if it is a voice in current debates, circumstances in the country, the type

of speech and the country’s political structure.60

Users are typically notified of the action taken in response to their content. But even this is

discretionary. Facebook provides that notice will not be provided if doing so would attract liability,

harm their community, or technical limitations restrict them from doing so.61 Harm to their

community could mean any number of things, as could a technical limitation.

57 Grimmelman op cit note 9 at page 67.
58 In this regard, see Facebook’s Transparency Center ‘How technology detects violations’, available at
https://transparency.fb.com/en-gb/enforcement/detecting-violations/technology-detects-violations/, accessed on 2
October 2022.
59 TikTok op cit note 46.
60 Facebook op cit note 45.
61 Facebook’s Terms of Service (last revised 26 July 2022), available at https://www.facebook.com/terms.php?ref=p,
accessed on 1 October 2022.

https://transparency.fb.com/en-gb/enforcement/detecting-violations/technology-detects-violations/
https://www.facebook.com/terms.php?ref=p

Importantly, social media platforms generally provide users with a review mechanism. This

enables a user to request a review in instances where their content has been incorrectly removed

or their account is disabled or deleted.62 Facebook has even appointed an Oversight Board – an

independent body that is mandated to hear select appeals.63

It is evident why such powers are akin to the exercise of public power. However, there is an

important distinction to highlight – platforms are not constrained in their exercise of power in the

same way that public bodies are. Their normative powers do not need to comply with a

constitution, which includes a Bill of Rights.64 Nor does such rule-making prescribe to the standard

principles of transparency, accountability and consistency.65 Their quasi-executive power is not

constitutionally required to be lawful, reasonable or procedurally fair.66 Their quasi-judicial power

is not informed by precedent or exercised by adjudicators who are bound by professional ethics

and principles of independence and impartiality.67

This important consideration sits amongst a myriad more, which are briefly detailed below.

(d) Considerations for content moderation

In this section, we explore four considerations that highlight the difficulties of content moderation

and the context within which it operates.

First, moderation is hard. In February 2022, the Russian government ordered Meta to stop fact-

checking and labelling content posted by Russian media sites.68 Meta refused, and Russia

announced it was restricting access to the platforms. Meta’s platforms were being used by Russian

citizens to express and organise themselves during a crisis.69 In essence, Meta had to decide

between allowing a government to disseminate any content it liked – possibly false – or keeping

62 Ibid.
63 See Facebook op cit note 58.
64 As is provided for in section 2 of the Constitution of the Republic of South Africa, 1996.
65 Emily B Laidlaw ‘Online platform responsibility and human rights’ in Luca Belli & Nicolo Zingales (eds)
Platform regulations (2017) 65 at 69.
66 As is provided for by section 33 of the Constitution.
67 As is provided for in section 165(4) of the Constitution.
68 Taylor Hatmaker ‘Russia Says it is Restricting Access to Facebook in the Country’ Techcrunch,25 February 2022
available at https://techcrunch.com/2022/02/25/russia-facebook-restricted-censorship-ukraine/, accessed on 3
March 2022.
69 Ibid.

https://techcrunch.com/2022/02/25/russia-facebook-restricted-censorship-ukraine/
https://techcrunch.com/2022/02/25/russia-facebook-restricted-censorship-ukraine/

its platforms accessible. An impossible choice. This demonstrates the difficult role that platforms

play in finding the appropriate balance between mitigating harm and enabling freedom of

expression. Such difficulties are heightened during a developing geopolitical crisis or global

pandemic.70

Second, platforms are influenced by multiple stakeholders, with various motives.71 They do not

exist in a vacuum and are subject to significant external pressures. Governments use the threat of

restriction or ban to force moderation in line with domestic agendas. This is evidenced by the

example above, as well as TikTok’s receipt of 4 156 government requests to restrict or remove

content in 2021.72 This is problematic when used improperly by governments to quell opposition

or suppress political criticism. It may be difficult for platforms to determine whether such requests

are justified or amount to censorship. This also applies to the user take-down process, where

abusive requests are made to silence other users.73

Third, moderation entails sifting through large quantities of information - 1.3 million posts are

shared every minute on Facebook.74 This has necessitated the use of automated technologies which

can sift through high volumes at speed. However, their use has resulted in additional challenges.

Their process, which removes content before it is published, creates a mechanism for prior

restraint, where technology makes snap decisions.75 The black-box nature of artificial intelligence

makes it difficult to understand why a decision was taken, and consequently how to challenge it.

This is underscored by critiques that the decisions are based on discriminatory assumptions and

biased datasets.76

Fourth, the globalised nature of the internet makes it difficult to shape rules that appropriately

account for local contexts.77 There is no universal definition of hate speech.78 Moderation policies

70 Evelyn Douek argues that the Covid-19 pandemic pushed content moderation into a state of emergency. See
Evelyn Douek ‘Governing online speech: from “posts-as-trumps” to proportionality and probability’ (2021) 121
Columbia Law Review 759 at 763.
71 Sander op cit note 32 at 959.
72 TikTok ‘Government Removal Requests Report’ 12 May 2022, available at
https://www.tiktok.com/transparency/en/government-removal-requests-2021-2/, accessed on 1 October 2022.
73 Keller op cite note 32 at 4.
74 Laidlaw op cit note 65 at 66.
75 Langvardt op cit note 28 at 1358.
76 Sander op cit note 31 at 958.
77 Balkin op cit note 37 at 2018.
78 Qwelane v South African Human Rights Commission 2021 (6) SA 579 (CC) para 79.

https://www.tiktok.com/transparency/en/government-removal-requests-2021-2/
https://www.tiktok.com/transparency/en/government-removal-requests-2021-2/

are either too permissive or too restrictive, depending on the history and context of a particular

jurisdiction.79 This also poses jurisdictional challenges for the vindication of rights.

These factors shape the reality that platforms will get it wrong. All moderators – human or artificial

– will make mistakes.80 But it is difficult to know whether a poor decision was a genuine mistake,

or whether it was dictated by ulterior motives. The next sections explore the impact that

moderation, and its mistakes, have on the right to freedom of expression in South Africa.

III. THE RIGHT TO FREEDOM OF EXPRESSION IN SOUTH AFRICA

(a) Un-packing the right to freedom of expression in South Africa

The right to freedom of expression is provided for in section 16 of the Constitution of South Africa.

It holds a significant space in our constitutional democracy because of our history of censorship.

That history entailed a regime of broad censorship where political, artistic and cultural expression

was restricted.81 Such a suppressive and intolerant past has garnered a strong appreciation for the

importance of the right.82

The rationales for the right have been acknowledged by our courts to include its role in the

functioning of democracy, in the search for truth and in every person’s ability to self-actualise.83

As noted by the Constitutional Court in Democratic Alliance v African National Congress:84

The Constitution recognises that people in our society must be able to hear, form and express

opinions freely. For freedom of expression is the cornerstone of democracy. It is valuable both

for its intrinsic importance and because it is instrumentally useful. It is useful in protecting

democracy, by informing citizens, encouraging debate and enabling folly and misgovernance

to be exposed. It also helps the search for truth by both individuals and society generally. If

society represses views it considers unacceptable, they may never be exposed as wrong. Open

79 Sander op cit note 31 at 956.
80 Douek op cit note 70 at 762.
81 Pierre De Vos ‘Rejecting the free marketplace of ideas: a value-based conception of the limits of free speech’
(2017) 33 SAJHR 359 at 361.
82 Economic Freedom Fighters v Minister of Justice and Correctional Services 2021 (2) SA 1 (CC) para 2.
83 South African National Defence Union v Minister of Defence and Another 1999 (4) SA 469 para 7.
84 2015 (2) SA 232 (CC) paras 122-3.

debate enhances truth-finding and enables us to scrutinise political argument and deliberate

social values.

The Constitutional Court has acknowledged a fourth rationale – ‘the encouragement of

tolerance.’85 Tolerance is considered a correlative of freedom of expression. It does not require the

acceptance of a particular view, but rather an acceptance that disagreements can be aired in public

and that unpopular views will not be silenced.86 This is underscored by the broad scope of the right

which protects unfavourable or controversial expressions too. As noted by the Constitutional Court

in Islamic Unity Convention v Independent Broadcasting Authority:87

Freedom of expression is applicable, not only to information or ideas that are favourably

received or regarded as inoffensive or as a matter of indifference, but also to those that offend,

shock or disturb the state or any sector of the population. Such are the demands of that

pluralism, tolerance and broadmindedness without which there is no democratic society.

The right to freedom of expression accordingly protects a wide range of expression, including that

which may be controversial, or offensive. It is broad in scope, and the Constitutional Court has

noted the obligation to interpret the scope of protected expression generously.88

(b) The operation and limitations of the right to freedom of expression

Despite its broad scope, the right is not absolute89 and ‘not all speech is created equal.’90 Our law

has placed lawful restrictions on the types of content that enjoy constitutional protection and

acknowledges instances where a limitation of the right is justified. This dichotomy is evident in

the construction of the right where section 16(1) broadly notes the scope of protection and section

16(2) specifies its limitations.

85 Qwelane supra note 78 para 69.
86 SANDU supra note 83 para 8.
87 2002 (4) SA 294 (CC) at para 26, which endorsed Handyside v the United Kingdom, no 5493/72 § 49, ECHR,
1976.
88 Laugh it off Promotions CC v South African Breweries International (Finance) BV 2006 (1) SA 144 (CC) para 47.
89 Iain Currie & Johan de Waal The Bill of Rights Handbook 5 ed (2005) at 163.
90 Dario Milo & Pamela Stein A Practical Guide to Media Law (2013) at 5.

Section 16(1) explicitly acknowledges certain important elements of the right including freedom

of the media,91 freedom to receive or impart information92; artistic creativity93 and freedom of

academic and scientific research.94

Section 16(2) lists three categories of expression that are not constitutionally protected. These are

propaganda for war;95 incitement of imminent violence;96 and advocacy of hatred that is based on

race, ethnicity, gender or religion and that constitutes incitement to cause harm.97 The exclusion

of some expression from constitutional protection was justified by the Constitutional Court in

Islamic Unity Convention v Independent Broadcasting Authority due to its potential to adversely

affect the dignity of others and to cause harm.98

In effect, this construction means that any regulation of the right which falls outside of section

16(2) and encroaches on the bounds of section 16(1), amounts to a limitation of the right.99 Such

a limitation is only constitutionally permissible if it is considered reasonable and justifiable in

terms of section 36(1) of the Constitution.100

South Africa’s jurisprudence contains many examples of justifiable limitations of the right to

freedom of expression. Legislative measures include the Promotion of Equality and Prevention of

Unfair Discrimination Act 4 of 2000. It is also lawfully curtailed by the exercise and enjoyment of

the rights to dignity and privacy.101 The reconciliation between such competing rights finds form

through the law of delict, specifically the actio iniuriarum.102

91 Section 16(1)(a) of the Constitution.
92 Section 16(1)(b) of the Constitution.
93 Section 16(1)(c) of the Constitution.
94 Section 16(1)(d) of the Constitution.
95 Section 16(2)(a) of the Constitution.
96 Section 16(2)(b) of the Constitution.
97 Section 16(2)(c) of the Constitution.
98 Islamic Unity supra note 87 para 30
99 Ibid para 34.
100 Currie op cit note 89 163.
101 Jonathan Burchell ‘The legal protection of privacy in South Africa: A transplantable hybrid’ (2009)13 Electronic
Journal of Comparative Law 1 at 1.
102 Debbie Collier ‘Freedom of expression in cyberspace: Real limits in a virtual domain’ (2005) 1 Stell LR, 21 at 28.

Platforms are important spaces for the exercise of the right to freedom of expression. Billions of

people around the world use them to share expression, and access information. They have

accordingly become regarded as the new marketplace for speech and ideas.103 In the American

case of Packingham v North Carolina,104 Justice Kennedy described social media platforms, as

the ‘modern public square’105 and noted that denying an individual access to them restricts their

ability to exercise their free speech rights. According to him, the internet, and platforms, are the

most important spaces for the exchange of ideas.106

This is true in South Africa too – there are currently over 41 million internet users.107 In 2022,

there were an estimated 28 million social media users which equates to nearly half the

population.108 A significant amount of our population is accordingly exercising their rights on the

internet and social media platforms specifically. Our courts have acknowledged this.

In Democratic Alliance v African National Congress109 the Constitutional Court included the

internet as a space for vigorous public debate, alongside public meetings, radio, television and

newspapers.110 In Afriforum v Pienaar,111 the Western Cape High Court extended the freedom of

the press and other media to social media platforms112 and in Tschilas v Touch Line Media the

court acknowledged the role that chat forums play in enabling debate,113 and focused on the

importance of freedom of expression on the internet.

Human rights apply online;114 implicating the considerations and limitations discussed. Online

expression can amount to hate speech and result in consequences in the same way that it could if

published in a newspaper. Our courts have navigated this by developing existing legal mechanisms

103 Klonick op cit note 22 at 1613.
104 137 S. Ct. 1730 (2017).
105 Ibid.
106 Ibid at 1735 (quoting Reno v ACLU, 521 U.S. 844, 868 (1997).
107 Datareportal ‘Digital 2022: South Africa’, available at https://datareportal.com/reports/digital-2022-south-
africa, accessed on 28 September 2022.
108 Ibid.
109 2015 (2) SA 232 (CC).
110 Ibid para 134.
111 2017 (1) SA 388 (WCC).
112 Ibid para 55.
113 Tsichlas v Touch Line Media (Pty) Ltd 2004 (2) SA 112 (W) at 119A-C.
114 Special Rapporteur op cit note 4 at 4.

https://datareportal.com/reports/digital-2022-south-africa
https://datareportal.com/reports/digital-2022-south-africa

to account for expression on the internet.115 This is appropriate for disputes that concern the

implications of one speaker’s expression on another’s rights; and where South African courts

ordinarily have jurisdiction.116 However, the structure of the regulation of expression has

changed.117 It doesn’t simply concern the state and the speaker. It concerns additional players

including platforms, internet service providers and trolls.118 Our courts have not adjudicated a

matter that concerns the implications of these new players’ actions on the freedom of expression

of a South African user. So what is their impact? The next section analyses how content moderation

undermines the right to freedom of expression.

IV. HOW CONTENT MODERATION UNDERMINES THE RIGHT TO FREEDOM OF

EXPRESSION IN SOUTH AFRICA

Content moderation undermines the right to freedom of expression in South Africa. This section

discusses how, by exploring four possible grounds of unconstitutionality.

Content moderation policies specify the rules that govern permissible expression on platforms.

These rules regulate expression. In South Africa, regulation is not required to give effect to the

right to freedom of expression.119 Because of this, any regulation of the right is considered a

limitation.120 Such regulation is not always fatal– it may be considered reasonable and justifiable

in terms of section 36 of the Constitution – and will accordingly stand. However, such an analysis

would fall short on the first hurdle every time because a moderation policy is not a law of general

application. The purpose of this section is not to determine the justifiability of each limitation but

rather to illustrate examples of how content moderation policies may undermine the right.

115 Daniel Sive & Alistair Price ‘Regulating expression on social media’ (2019) 136 SALJ 51 at 67. See for
example, The Chinese Association Gauteng v Henning and Others (EQ2/2017) [2022] ZAGPHJC 590 (28 July
2022), which concerned expression on Facebook, some of which amounted to hate speech and was found to
contravene the Equality Act.
116 In the matter of Chinese Association supra note 115, for example, all of the parties fell within the jurisdiction of
the court.
117 Jack M Balkin conceives of this new structure as a triangle. See Balkin op cit note 37 at 2015.
118 Ibid.
119 Print Media South Africa v Minister of Home Affairs 2012 (6) SA 443 (CC) para 51.
120 Ibid para 51. However, it must be noted that regulation of the types of expression provided for in section 16(2) of
the Constitution would not be considered a limitation of the right in section 16(1), see Qwelane supra note 78 para
76.

They may do so, for four reasons. First, they prohibit categories of protected expression that are

not included in section 16(2) of the Constitution. Second, they are vague and fall short of the

requirements of the rule of law.121 Third, the process of moderation entails prior restraint and

fourth, their application extends to the private sphere.

(a) The limitations extend beyond section 16(2)

Facebook, TikTok and Twitter all place limitations on categories of expression that would

ordinarily be constitutionally protected. Facebook includes limitations on expression concerning

adult nudity and sexual activity,122 and suicide and self-injury.123 TikTok limits ‘disordered eating’

and violent and graphic content and Twitter restricts content that suppresses participation in civic

processes.124 All three platforms prohibit or place limitations on false information. 125 126

These categories of expression are not listed in section 16(2) of the Constitution and are

accordingly considered protected expressions in terms of section 16(1). Their restriction amounts

to a limitation of the right.

(b) The limitations are vague

Section 1(c) of the Constitution provides that South Africa is founded on the rule of law. This

requires that rules must be sufficiently clear to enable individuals to comply. In this regard, the

Constitutional Court has held:

[The rule of law] requires that laws must be written in a clear and accessible manner. What is

required is reasonable certainty and not perfect lucidity. The doctrine of vagueness does not

require absolute certainty of laws. The law must indicate with reasonable certainty to those

who are bound by it what is required of them so that they may regulate their conduct

accordingly.127

121 As is provided for in section 1(c) of the Constitution.
122 Facebook op cit note 45.
123 Ibid.
124 The Twitter Rules, available at https://help.twitter.com/en/rules-and-policies/twitter-rules, accessed on 28
September 2022.
125 Facebook regulates ‘misinformation’, see Facebook op cit note 45.
126 TikTok regulates ‘harmful misinformation’ and ‘integrity and authenticity’, see TikTik op cit note 46.
127 Affordable Medicines Trust v Minister of Health 2006 (3) SA 247 (CC) para 108.

https://help.twitter.com/en/rules-and-policies/twitter-rules

The Court has further provided that for a rule to be unconstitutionally vague – it must be more than

just poorly constructed – it must be completely ‘meaningless and unworkable.’128TikTok regulates

misinformation by providing:

[m]isinformation is defined as content that is inaccurate or false. We will remove

misinformation that causes significant harm to individuals, our community, or the larger

public regardless of intent.129

This limitation is inherently vague – what is false? Such a classification necessarily entails

universal truth and agreement on such truth.130 Facebook acknowledges the impossibility of this

standard by noting that ‘what is true one minute may not be true the next minute.’131 This is of

particular concern during developing or unprecedented events where there are no previous facts to

verify content132 or where there are no authoritative sources to defer to.133 The wording poses the

following difficulties - would misleading information (which is not false) be considered

inaccurate? Would genuine opinion, which is identified as such, fall within the ambit of this section

if those making determinations disagreed with it? What happens when authoritative sources

contradict each other?134

As evidenced by these concerns, the wording of this limitation is not sufficiently clear to enable a

user to know, with certainty, whether their expression falls foul of the provision. Vague and overly

broad provisions, of which there are many, limit the right to freedom of expression and may result

in users' self-censoring, which has a ‘chilling effect’ on the right to freedom of expression.135

The operation of content moderation may, in some instances, entail the censorship of content

before it is published on a platform. This process amounts to a system of prior restraint.136 It occurs

when content is automatically flagged and never published. Prior restraints have been criticised

128 National Credit Regulator v Opperman 2013 (2) SA 1 (CC) para 46.
129 TikTok op cit note 46.
130 Molina op cit note 18 at 183.
131 Facebook op cit note 45.
132 Molina op cit note 18 at 183.
133 Douek op cit note 70 at 762.
134 Ibid.
135 Print Media supra note 119 para 23
136 Langvardt op cit note 28 at 1359.

for undermining the right to freedom of expression and making expression beholden to the

government or an administrative body.137 It is contended that such mechanisms cause self-

censorship, have a chilling effect on freedom of expression and undermine public

discourse.138Because of this, our courts have adopted a cautious approach to prior restraint.139

In Print Media South Africa v Minister of Home Affairs140, the Constitutional Court held that such

a mechanism unjustifiably limited the right to freedom of expression and was unconstitutional. In

making such a determination, the court noted that the nature of prior restraint entails a shift of

control from the right bearer to an administrative body.141 It noted:

By investing an administrative body with the exclusive power to grant permission to publish

certain material, as well as the power to punish for denying it the opportunity to do so, what is

engendered is a scheme in which expression must be justified before, and as a necessary

condition for, its release into the public realm.142

Importantly, this type of restraint excludes the possibility of scrutiny and engagement from the

public and is likely to lead to over-restriction of content.143 Publications are more likely to be

restricted before publication than after, and bodies that are obligated and incentivised to classify

and restrict content are more likely to do so.144 Our courts accordingly only permit it in very narrow

circumstances.145 Such a scheme used by platforms may undermine the right to freedom of

expression and be considered unconstitutional.

(d) Restrictions apply to private expression

137 George Devenish ‘Prior judicial restraint and media freedom in South Africa – Some cause for concern’ (2011)
74 Journal of Contemporary Roman-Dutch Law 12 at 13.
138 Ibid.
139 See Government of the Republic of South Africa v Sunday Times Newspaper 1995 (2) SA 221 (T).
140 Supra note 119 para 51.
141 Print Media supra note 119 para 58.
142 Ibid.
143 Ibid para 59.
144 Ibid.
145 Midi Television (Pty) Ltd v Director of Public Prosecutions 2007 (5) SA 540 (SCA) para 15.

The content moderation rules of some platforms apply to private messages.146 This is arguably

necessary to guard against targeted harassment or hate speech. However, the Constitutional Court

has held that even lawful prohibitions on hate speech do not permeate private conversations.147

Providing, that extending them to the private realm is not consistent with the purpose of such

prohibitions which are aimed at limiting public expressions of hate speech.148 Accordingly, the

operation of content moderation policies to private messages may be found to unjustifiably limit

the right to freedom of expression.

V. THE ROLE OF THE STATE

(a) Why involve the state?

Social media platforms are owned by private, international companies.149 They develop the rules

of engagement for access to their platforms which are accepted by users, who are also private

actors. Such is the legal, contractual arrangement governing access to and use of social media

platforms. Private international companies are not enjoined to protect and promote the rights in

the South African Bill of Rights.150 So why does this relationship invoke the role of the state at

all?

The answer is that it doesn’t have to, but it should. States all over the world are waking up to the

fact that content moderation shapes public discourse and impacts human rights.151 What to censor

is an important public policy consideration, and in determining that, platforms exercise powers

akin to public or state power.152 Because of this, it is contended that states have a duty to constrain

the exercise of such power153 and there have been calls for public scrutiny to guard against

146 Twitter notes that its policies apply to direct messages, which are private. See in this regard Twitter’s Help
Center ‘Our range of enforcement options’ available at https://help.twitter.com/en/rules-and-policies/enforcement-
options, accessed on 2 October 2022.
147 Qwelane supra note 78 para 118.
148 Ibid.
149 For example, Meta is a private company, incorporated in the USA, that owns Facebook, Instagram and
Whatsapp, amongst others.
150 Kyle Langvardt remarks that ‘[t]hat Facebook is not a governmental actor, of course, relieves all formal
constitutional concerns about the company’s content restriction policies.’ See Langvardt op cit note 28 at 1357. See
also Kaunda v President of the Republic of South Africa 2004 (10) BCLR 1009 (CC) para 37 which held that the
Bill of Rights does not have general application outside South Africa.
151 Ibid.
152 Ibid.
153 Belli op cit note 41 at 60.

https://help.twitter.com/en/rules-and-policies/enforcement-options
https://help.twitter.com/en/rules-and-policies/enforcement-options

abuse.154 Lucca, for example, remarked that ‘it is simply unacceptable for States to throw their

hands up and let platforms define the content, scope and limitations of fundamental rights without

adequate constraints.’155

There is accordingly a growing shift away from the view that the private nature of social media

companies insulates them from a duty to respect and protect the rights of their users.156

International proponents contend that such a view is supported by the Human Rights Council’s

adoption of the United Nations’ Guiding Principles on Business and Human Rights157 which

provides a framework for private companies to respect and protect human rights. Further,

international human rights law requires states to ensure that human rights violations are not

committed by private parties.158 Specifically, the state’s obligation to respect, protect and promote

human rights extends to the oversight of private companies.159 This view is supported by the

Council of Europe which recently remarked that placing these considerations squarely within the

realm of contract would be ‘untenable’ considering jurisprudence from the European Court of

Human Rights.160

What this jurisprudence shows is an acknowledgement that the internet and platforms are vital for

the exercise of human rights, specifically freedom of expression.161 Their governance cannot be

viewed only in terms of contract. It also shows a reimagining of the right to freedom of expression

to include positive obligations. Such obligations require states to take certain domestic measures

to ensure the right may be exercised and is not interfered with by private, international persons.162

In Dink v Turkey163 the court pronounced that states have a duty to ensure favourable conditions

154 Ibid.
155 Ibid.
156 Evelyn Douek ‘The limits of international law in content moderation’ (2021) 6 UC Irvine Journal of
International, Transnational and Comparative law 37 at 40.
157 Human Rights Council Resolution 17/4 United Nations Document A/HRC/Res/17/4 (6 July 2011); John Ruggie
(Special Representative of the Secretary-General), Guiding Principles on Business and Human Rights:
Implementing the United Nations ‘Protect, Respect and Remedy’ Framework, UN Doc A/HRC/17/31 (21 March
2011).
158 Aleksandra Kuczerawy ‘The power of positive thinking: Intermediary liability and the effective enjoyment of the
right to freedom of expression’ (2017) KU Leuven Centre for IT & IP Law CiTiP Working Paper 30/2017 at 6.
159 United Nation’s General Comment No 31 [80] adopted on 29 March 2004, available at
https://www.refworld.org/docid/478b26ae2.html, accessed on 19 October 2022.
160 Council of Europe ‘Guidance note on content moderation’ (2021) adopted by the Steering Committee for Media
and Information Society Content Moderation at 21.
161 See Cengiz v Turkey, nos 48226/10 and 14027/11 (2015) para 49.
162 Kuczerway op cit note 159 at 8.
163 Dink v Turkey (2010) ECtHR.

https://www.refworld.org/docid/478b26ae2.html

exist to enable everyone to exercise their right to freedom of expression. In Özgür Gündem v

Turkey, the European Court of Human Rights held that the exercise of the right ‘does not depend

merely on the State’s duty not to interfere but may require positive measures of protection.’164 This

reimagining enables individuals to hold the state accountable if it has not taken appropriate

domestic measures to constrain the powers of private, international companies. Given the impact

of content moderation, such inadequate measures would undermine their duty to respect, protect

and promote the right to freedom of expression.

(b) Reconceptualising the right in South Africa

Such a reconceptualisation is supported in South Africa’s jurisprudence. The right is traditionally

considered a negative one; requiring the state to not interfere with it.165 However, our Courts have

acknowledged that even civil and political rights may confer a positive duty.166 Such argument is

premised on section 7(2) of the Constitution which enjoins the state to ‘respect, protect, promote

and fulfil the rights in the Bill of Rights.’167 This was recognised by the technical committee

working on the Constitution in 1996. In a commentary on section 7 of the Constitution, it noted:168

[t]the obligation ‘to protect’ the rights requires States to take positive steps to prevent a right

from being infringed by both State and private actors. The obligation ‘to promote’ the rights

refers to the duty of the State to take steps to create the necessary culture and social conditions

in which the full enjoyment of human rights is possible.

There is accordingly a duty on the state to ensure that rights are not infringed by private actors.

Further, states must take necessary measures to ensure that rights may be exercised and enjoyed.169

This means that the state should use its resources and powers to ensure that the conditions exist

for freedom of expression to be exercised. This would include its exercise online. Such positive

164 Özgür Gündem v Turkey (2000) ECtHR para 43.
165 Currie op cit note 89 at 571.
166 See August v Electoral Commission 1999 (3) SA 1 (CC), which considers the positive obligations enjoined by the
right to vote.
167 See Minister of Health v Treatment Action Campaign 2002 (5) SA 721 (CC) at para 39
168 Memorandum of the Panel of Constitutional Experts and Technical Committee 4, 8 March 1997, available at
https://www.justice.gov.za/legislation/constitution/history/LEGAL/CP008036.PDF, accessed on 22 October 2022.
169 De Vos op cit note 81 at 378.

https://www.justice.gov.za/legislation/constitution/history/LEGAL/CP008036.PDF

measures could include the adoption of appropriate administrative or legislative measures to

achieve this.170

Even a strictly contractual or private nature of platforms may not frustrate such an approach. Our

courts have demonstrated a willingness to intervene in private, contractual matters in pursuit of

principles of reasonableness, fairness and justice. This is evident in Barkhuizen v Napier171 where

the court held that freedom of contract may be constrained if its enforcement would undermine

such principles. The Constitutional Court noted that to hold otherwise – that a court could never

intervene – would tie the hands of justice, which could never be the case.172 Public policy is now

dictated by the Constitution173 and the Supreme Court of Appeal has acknowledged that a

contractual term that limits a constitutional value will not be enforced.174 Our courts have also

intervened in procedurally unfair private, disciplinary matters.175 It has done so for churches176

and professional associations.177 Their intervention has been rationalised on the importance of the

consequences – it may cause reputational damage or economic hardship.178

The global shift in how moderation and the right are conceptualised is mirrored by shifts in

approaches to regulation. Content moderation used to be predominantly governed by self-

regulation, with social media companies voluntarily determining and enforcing their own

governance structures, through their moderation policies.179 However, states have become

increasingly more involved in the governance of content moderation.180 They have begun to take

positive steps to constrain the power of platforms to protect and promote the right to freedom of

expression. States have taken different approaches to regulation and South Africa’s approach is

detailed below.

170 Ibid.
171 Barkhuizen v Napier 2007 (5) SA 323 (CC) para 73.
172 Ibid.
173 Sive op cit note 115 at 75.
174 Bredenkamp v Standard Bank of SA Ltd 2010 (4) SA 468 (SCA) para 50.
175 Cora Hoexter Administrative Law in South Africa 2 ed (2012) at 444.
176 Ndara v Umtata Presbytery, Nederduitse Gereformeerde Kerk in Afrika (Transkei) 1990 (4) SA 22 (Tk).
177 Jockey Club v Forbes 1993 (1) SA 649 (A).
178 Hoexter op cit note 176 at 444.
179 Christopher Marsden Internet Co-regulation: European Law, Regulatory Governance and Legitimacy in
Cyberspace 1 ed (2011) at 212.
180 Ibid.

VI. SOUTH AFRICA’S REGULATION OF ONLINE EXPRESSION

This section details the regulatory responses that South Africa has taken to freedom of expression

online. Such regulation is provided for in the Electronic Communications and Transactions Act 25

of 2002 (‘ECTA’) and the Films and Publications Act 65 of 1996 (‘FPA’). The FPA has recently

been amended181 to bring online content into its remit.182 The amendments include several vague

provisions which pose challenges for its implementation and have drawn significant criticism.183

Together, these laws do four things: first, provide conditional limitation of civil liability for

platforms. Second, establish a parallel take-down process that enables the removal of content

online. Third, establish a complaints mechanism that may result in the removal of content and

fourth, impose a classification scheme for online content that amounts to prior restraint.

(a) Conditional limitation of liability

Provision is made to exclude platforms from civil liability for content on their platform. They

would accordingly not be held liable for defamation, for example, by hosting or publishing

defamatory content. Such a regulatory position acknowledges that platforms have limited control

over the content users publish.184 It is considered necessary to incentivise platforms to retain their

service offerings which enables the exercise of freedom of expression.185

Such limitation of liability is provided for in chapter XI of ECTA. However, liability is not

automatically excluded. It only applies if certain conditions are met. First, the service provider

must be a member of a recognised representative body.186 To be recognised, members must be

subject to a code of conduct that includes adequate standards,187 there must be requirements for

membership and the representative body must be capable of enforcing compliance.188 Second, it

181 It was amended by the Films and Publications Amendment Act no 19 of 2020, which came into effect on 1
March 2022.
182 See the definition of ‘publication’ in section 1 which explicitly includes expressions published using the internet.
183 See, for example, Noxolo Majavu & Palesa Dlamini ‘Films and Publications Amendment Act leaves online
content producers hot under the collar’ News24 5 March 2022, available at
https://www.news24.com/citypress/news/films-and-publications-amendment-act-leaves-producers-hot-under-the-
collar-20220305, accessed on 10 October 2022.
184 Sive op cit note 115 at 70.
185 Keller op cit note 32 at 9.
186 Section 72(a) of ECTA.
187 Section 71(2)(a) and (c) of ECTA.
188 Section 71(2)(d) of ECTA.

https://www.news24.com/citypress/news/films-and-publications-amendment-act-leaves-producers-hot-under-the-collar-20220305
https://www.news24.com/citypress/news/films-and-publications-amendment-act-leaves-producers-hot-under-the-collar-20220305

must have adopted and implemented its official code of conduct.189 To date, no social media

platforms have met these conditions.190 It is not clear whether social media platforms would

collectively associate, or agree on a code of conduct.

Additional requirements for immunity apply depending on the operations of the service provider.

Social media platforms typically store data, such as a tweet or post, which is developed and shared

by a recipient or user of their service. Platforms would accordingly be considered a ‘host’ in terms

of ECTA.191 For liability to be excluded, hosts must meet three additional requirements. They must

not have actual knowledge that the content infringes the rights of someone else192 or knowledge

of the facts or circumstances of the rights infringement,193 and they must respond to a take-down

notice ‘expeditiously.’194

(b) The take-down notification process

The take-down notification process is provided for in section 77 of ECTA. Any person is

competent to file a notification ‘of unlawful activity’ with the platform. This legislatively enables

private individuals to interfere with another person’s expression.195 Service providers are required

to expeditiously remove or disable access to the data following receipt of the take-down

notification.196 A platform will not be held liable if it incorrectly removes content in response to a

take-down notification.197

The limitation of liability scheme and the take-down process provided for in ECTA are

problematic for two reasons – they incentivise removal and outsource important decisions about

the exercise of rights.

189 Section 72(b) of ECTA.
190 Department of Communications GN 588 GG32252 of 22 May 2009.
191 Section 75(1) of ECTA considers a host to be a service provider that ‘provides a service that consists of the
storage of data provided by a recipient of the service.’ Price and Sive concur in the assertion that a social media
platform would be considered a ‘host’ for the purposes of ECTA, see Sive op cit note 115 at 71.
192 Section 75(1)(a) of ECTA.
193 Section 75(1)(b) of ECTA.
194 Section 75(1)(c) of ECTA.
195 Kuczerawy op cit note 159 at 6.
196 Section 75(1)(c) of ECTA.
197 Section 77(3) of ECTA.

(i) Incentivising removal

ECTA incentivises platforms to remove content by making the exemption of liability conditional

upon an expeditious response to a take-down notification.198 To avoid liability, a platform will

likely act cautiously and remove the content, instead of weighing up any conflicting rights.199 Such

an exercise is burdensome and time-consuming,200 and they are incentivised by ECTA to respond

quickly. If they are slow or do not respond at all, the limitation of liability may not apply to them.

201 To avoid legal and financial risk, a platform is likely to just remove the impugned content,

regardless of its legality.202 Further, if a domestic jurisdiction has legally mandated a notice and

takedown system it ‘stacks the deck in favour of the accusers.’203 Platforms are likely to take the

request at face value, instead of determining whether the request is legitimate.204 This is bolstered

by the fact that a platform will not be held liable if it incorrectly removes content in response to a

take-down notification.205 In real terms, this means that when confronted with difficult or ‘grey-

area’ expressions such as satire, humour or a developing controversial debate, platforms will

simply remove it.206 Such an approach does not interpret protected expression broadly, or tolerate

controversial or offensive expression, as required by our Constitution.207 This may very well lead

to over-removal and the censorship of lawful expression online which undermines the right to

freedom of expression.208

(ii) Outsourcing rights-based determinations

Such a take-down notification scheme is considered an ‘inappropriate transfer of juridical authority

to the private sector.’209 ECTA does not expressly provide which content may be lawfully

198 Kuczerawy op cit note 159 at 4.
199 Ibid at 5.
200 Belli op cit note 41 at 52.
201 Section 75(1)(c).
202 Kuczerawy op cit note 159 at 5.
203 Keller op cit note 32 at 5.
204 Ibid.
205 Section 77(3) of ECTA.
206 Keller op cit note 32 at 5.
207 Laugh it off supra note 88 para 47.
208 Sive op cit note 115 at 71.
209 European Commission, Summary of the results of the Public Consultation on the future of electronic commerce
in the Internal Market and the Implementation of the Directive on electronic commerce (2000/31/EC), available at
http://ec.europa.eu/internal_market/consultations/docs/2010/ecommerce/summary_report_en.pdf, accessed on 18
October 2022.

http://ec.europa.eu/internal_market/consultations/docs/2010/ecommerce/summary_report_en.pdf

removed, it simply requires that the complainant identify the right that has been allegedly infringed

and the content that is the subject of unlawful activity.210 Defamatory content may infringe on the

right to dignity, but it may also be lawful if it is true and in the public interest.211 Rights

infringements that occur because of expression generally entail consideration of the two opposing

rights at play. It is not enough to simply allege the existence of a rights infringement to justify the

removal of content. As noted above, ECTA doesn’t incentivise careful weighing up.

ECTA does not require a platform to hear from the content creator or adhere to the principle of

audi alteram partem in any other way. Nor does it place any transparency or accountability

requirements on the platform to enable a state body to oversee the implementation and outcomes

of the process. Further, ECTA does not establish a review mechanism, and there are accordingly

no due process requirements included in the law, despite its transfer of juridical authority.

Following a platform’s internal review process will likely result in the platform simply advising

that they were complying with domestic law.

Section 18E of the FPA establishes a complaints mechanism that allows any person to lodge a

complaint with the Film and Publication Board (‘Board’) about online content that is unclassified,

prohibited or potentially prohibited.212 The content that is prohibited includes propaganda for

war;213 content that incites imminent violence;214 or ‘advocates hatred based on any identifiable

group characteristic and that constitutes incitement to cause harm and imminent violence.’215

These prohibitions largely mirror the categories of unprotected expression listed in section 16(2)

of the Constitution.216 It also includes child pornography.217 If the Board determines the content is

unclassified or constitutes prohibited content, it may issue a take-down notice to a platform as

210 Sections 77(1)(c) and (d) of ECTA.
211 See Times Media Ltd and Others v Nisselow [2005] 1 All SA 567 (SCA).
212 Section 18E(1) of the FPA.
213 Section 16(2)(b) of the FPA.
214 Section 16(2)(c) of the FPA.
215 Section 16(2)(d) of the FPA.
216 However, notably, the prohibition of the advocacy of hatred extends beyond the scope provided for in section
16(2)(c) of the Constitution by extending the listed grounds and requiring the incitement of harm and imminent
violence.
217 Section 1 definition of ‘prohibited content’ read with section 16(4) and section 18(3) of the FPA.

provided for in ECTA.218 As detailed above, the platform is likely to respond by removing the

impugned expression, without considering the legality of the request.

This complaints mechanism expressly applies to online content219 and poses two concerns. First,

it outsources important juridical decisions to an administrative body. Effectively, the board is

empowered to determine whether expression advocates hatred or incites imminent violence. As

detailed above, these are difficult decisions and our courts have taken issue with administrative

bodies making such decisions which have huge implications for human rights.220

Second, a huge amount of content is posted online. It is unclear whether the Board has sufficient

resources and capacity to deal with such complaints effectively. This is important to consider in

light of the need to remove some content quickly to resolve continuing harm.

(d) The obligation to classify

Section 16(2) of the FPA requires that publications, which include ‘any content made available

using the internet,’221 be submitted to the Board for classification before they are distributed. To

distribute explicitly includes ‘to stream content through the internet, social media or other

electronic mediums.’222 Content published online accordingly falls within the scope of this

obligation.

In a media briefing, it was noted that ‘non-commercial online distributors’ are not required to

submit their content for classification before posting it online.223 This would exclude personal

users of social media platforms. However, it is not explicitly provided for in the FPA or the

regulations and it is not immediately clear whether someone who posts on social media for private,

income generation purposes, such as an ‘influencer’ would be considered a ‘commercial online

distributor’ or not.

218 Section 18E(2) of the FPA.
219 Section 18E of ECTA provides that any person may complain about such content ‘in relation to services being
offered online by any person.’
220 See Print Media supra note 119.
221 Definition of ‘publication’ in section 1 of the FPA.
222 Definition of ‘distribute’ in section 1 of the FPA.
223 Film and Publication Amendment Act media briefing, streamed live on 3 March 2022, available at
https://www.youtube.com/watch?v=SSpSEqw5HYw, accessed on 20 October 2022.

https://www.youtube.com/watch?v=SSpSEqw5HYw

The consequence of classification is that the content will either be classified as a refused

classification,224 will be restricted225 or ascribed an age restriction.226 Content classified as ‘refused

classification’ and ‘XX’ cannot be broadcast or distributed, and it is an offence to do so.227

In effect, these sections create a mechanism for administrative prior restraint, a process that enables

the restriction of content before it sees the light of day.228 As discussed above, our courts are

hesitant about such mechanisms noting that ‘the free flow of constitutionally protected expression

is the rule and administrative prior classification should be the exception.’229 These provisions

may accordingly be subject to constitutional challenge.

Regardless, the uncertainty about the application of this section to ordinary non-commercial users

of social media may have a chilling effect on users, undermining their exercise of freedom of

expression online.

South Africa’s regulatory approach to online expression raises several concerns – it incentivises

over-removal, does not require adherence to any due-process principles, introduces a mechanism

for administrative prior restraint, poses practical challenges and may chill expression. I submit that

such an approach does not extinguish the state’s constitutional obligation to respect, protect and

promote the right to freedom of expression.230 In fact, in several ways, it undermines it.

VII. CONCLUSION

The exercise of freedom of expression has been fundamentally changed by the internet. This has

reshaped the structure of its regulation – requiring consideration of the role of new, private bodies.

A role which has enabled expression, but also undermined it. The consequences of content

moderation are enormous and its impact on human rights is significant. This poses a regulatory

challenge for states all over the world. Some have risen to the challenge by reimagining freedom

of expression to place positive duties on the state to intervene – requiring the enactment of

224 Section 16(4)(a) of the FPA.
225 Section 16(4)(b) of the FPA.
226 Section 16(4)(d) of the FPA.
227 Section 24A(2) of the FPA.
228 Devenish op cit note 138 at 13.
229 Print Media supra note 119 para 52.
230 Section 7(2) of the Constitution.

measures to protect it online. Our current legislative measures do not give affect to the right to

freedom of expression. They cannot be said to do so when they incentivise removal and have a

chilling effect on expression. Our jurisprudence, however, includes approaches which may be

developed to shift our conception of the right to freedom of expression. We may yet rise to the

challenge.

Bibliography

Cases

1. Affordable Medicines Trust v Minister of Health 2006 (3) SA 247 (CC).

2. Afriforum NPC & Another v Pienaar 2017 (1) SA 388 (WCC).

3. August v Electoral Commission 1999 (3) SA 1 (CC).

4. Barkhuizen v Napier 2007 (5) SA 323 (CC).

5. Bredenkamp v Standard Bank of SA Ltd 2010 (4) SA 468 (SCA).

6. Cengiz v Turkey, nos. 48226/10 and 14027/11 (2015).

7. Democratic Alliance v African National Congress 2015 (2) SA 232 (CC) paras 122-3.

8. Dink v Turkey (2010) ECtHR.

9. Economic Freedom Fighters v Minister of Justice and Correctional Services 2021 (2) SA 1

(CC).

10. Government of the Republic of South Africa v Sunday Times Newspaper 1995 (2) SA 221

(T).

11. Islamic Unity Convention v Independent Broadcasting Authority and Others 2002 (4) SA

294.

12. Jockey Club v Forbes 1993 (1) SA 649 (A).

13. Laugh It Off Promotions VV v SAB International (Finance) BV t/a Sabmark International

2005 (8) BCLR 743 (CC).

14. Midi Television (Pty) Ltd v Director of Public Prosecutions 2007 (5) SA 540 (SCA).

15. Minister of Health v Treatment Action Campaign 2002 (5) SA 721 (CC).

16. National Credit Regulator v Opperman 2013 (2) SA 1 (CC).

17. Ndara v Umtata Presbytery, Nederduitse Gereformeerde Kerk in Afrika (Transkei) 1990

(4) SA 22 (Tk).

18. Packingham v North Carolina 137 S. Ct. 1730 (2017).

19. Print Media South Africa v Minister of Home Affairs 2012 (6) SA 443 (CC).

20. Qwelane v South African Human Rights Commission 2021 (6) SA 579 (CC).

21. Özgür Gündem v Turkey (2000) ECtHR.

22. Reno v ACLU, 521 U.S. 844, 868 (1997).

23. South African National Defence Union v Minister of Defence and Another 1999 (4) SA

469.

24. The Chinese Association Gauteng v Henning and Others (EQ2/2017) [2022] ZAGPHJC

590 (28 July 2022).

25. Tsichlas v Touch Line Media (Pty) Ltd 2004 (2) SA 112 (W) at 119A-C.

Books

26. Christopher Marsden Internet Co-regulation: European Law, Regulatory Governance and

Legitimacy in Cyberspace 1 ed (2011).

27. Cora Hoexter Administrative Law in South Africa 2 ed (2012).

28. Dario Milo & Pamela Stein A Practical Guide to Media Law (2013).

29. Emily B Laidlaw ‘Online platform responsibility and human rights’ in Luca Belli &and

Nicolo Zingales (eds) Platform regulations (2017) 65.

30. Iain Currie & Johan de Waal The Bill of Rights Handbook 5 ed (2005).

31. Tarleton Gillespie Custodians of the internet: Platforms, content moderation, and the

hidden decisions that shape social media (2018)

32. Luca Belli, Pedro Augusto Francisco & Nicolo Zingales ‘Law of the land or law of the

platform? Beware of the privatisation of regulation and police’ in Luca Belli & Nicolo

Zingales (eds) Platform regulations (2017) 41.

Journals

33. Aleksandra Kuczerawy ‘The power of positive thinking: Intermediary liability and the

effective enjoyment of the right to freedom of expression’ (2017) KU Leuven Centre for IT

& IP Law CiTiP Working Paper 30/2017.

34. Alexandre Bovet & Hernan Makse ‘Influence of fake news in Twitter during the 2016 US

presidential election’ (2019) 10 Nature Communications 1.

35. Alvaro Figueira & Luciana Oliveria ‘The current state of fake news: challenges and

opportunities’ (2017) 121 Procedia Computer Science 180.

36. Barrie Sander ‘Freedom of expression in the age of online platforms: the promise and

pitfalls of a human rights-based approach to content moderation’ (2020) 43 Fordham

International Law Journal 939

37. Daniel Sive & Alistair Price ‘Regulating expression on social media’ (2019) 136 SALJ 51.

38. Daphne Keller ‘Internet platforms: observations on speech, danger, and money’ (2018)

Hoover Working Group on National Security, Technology and Law, Aegis Series Paper

No. 1807

39. Debbie Collier ‘Freedom of expression in cyberspace: real limits in a virtual domain’

(2005) 1 Stell LR 21.

40. Evelyn Douek ‘Governing online speech: from “posts-as-trumps” to proportionality and

probability’ (2021) 121 Columbia Law Review 759.

41. Evelyn Douek ‘The limits of international law in content moderation’ (2021) 6 UC Irvine

Journal of International, Transnational and Comparative law 37.

42. Fawzia Cassim ‘Regulating hate speech and freedom of expression on the internet:

promoting tolerance and diversity’ (2015) 28(3) South African Journal of Criminal Justice

303.

43. George Devenish ‘Prior judicial restraint and media freedom in South Africa – some cause

for concern’ (2011) 74 Journal of Contemporary Roman-Dutch Law 12.

44. Jack M. Balkin ‘Free Speech is a Triangle’ (2018) 118 Columbia Law Review 2011.

45. James Grimmelmann ‘The virtues of moderation’ (2015) 17 Yale Journal of Law and

Technology 42.

46. Jonathan Burchell ‘The legal protection of privacy in South Africa: a transplantable

hybrid’ (2009) 13 Electronic Journal of Comparative Law, 1.

47. Kate Klonick ‘The new governors: the people, rules and processes governing online

speech’ (2018) 131 Harvard Law Review 1598.

48. Kyle Langvardt ‘Regulating online content moderation’ (2018) 106 Georgetown Law

Journal 1353.

49. Lior Strahilevitz ‘Wealth without markets?’ (2007) 116 Yale Law Journal 1472.

50. Maria D Molina, S. Shyam Sundar, Thai Le & Donwon Lee ‘“Fake News” is not simply

false information: a concept explication and taxonomy of online content’ (2021) 65(2)

American Behavioural Scientist 180.

51. Pierre De Vos ‘Rejecting the free marketplace of ideas: a value-based conception of the

limits of free speech’ (2017) 33 SAJHR 359.

52. Tarleton Gillespie, ‘Platforms are not intermediaries’ (2018) 2.2 Georgetown Law

Technology Review 198.

Reports

53. Council of Europe ‘Guidance note on content moderation’ (2021) adopted by the Steering

Committee for Media and Information Society content moderation.

54. European Commission, Summary of the results of the Public Consultation on the future of

electronic commerce in the Internal Market and the Implementation of the Directive on

electronic commerce (2000/31/EC), available at

http://ec.europa.eu/internal_market/consultations/docs/2010/ecommerce/summary_report_

en.pdf, accessed on 18 October 2022.

55. Human Rights Council Resolution 17/4 United Nations Document A/HRC/Res/17/4 (6

July 2011); John Ruggie (Special Representative of the Secretary-General), Guiding

Principles on Business and Human Rights: Implementing the United Nations ‘Protect,

Respect and Remedy’ Framework, UN Doc A/HRC/17/31 (21 March 2011).

56. Report of the Special Rapporteur on the Promotion and Protection of the Right to Freedom

of Opinion and Expression Doc ‘A/HRC/17/27’ 16 May 2011.

57. Report of the Special Rapporteur on the Promotion and Protection of the Right to Freedom

of Opinion and Expression Doc ‘A/HRC/32/38’ 11 May 2011.

58. United Nation’s General Comment No 31 [80] adopted on 29 March 2004, available at

https://www.refworld.org/docid/478b26ae2.html, accessed on 19 October 2022.

59. Memorandum of the Panel of Constitutional Experts and Technical Committee 4, 8 March

1997, available at

https://www.justice.gov.za/legislation/constitution/history/LEGAL/CP008036.PDF,

accessed on 22 October 2022.

http://ec.europa.eu/internal_market/consultations/docs/2010/ecommerce/summary_report_en.pdf
http://ec.europa.eu/internal_market/consultations/docs/2010/ecommerce/summary_report_en.pdf
https://www.refworld.org/docid/478b26ae2.html
https://www.justice.gov.za/legislation/constitution/history/LEGAL/CP008036.PDF

Newspaper articles

60. David Gilbert ‘Facebook and Instagram are censoring protests against police violence in

Nigeria’ Vice 21 October 2021, available at

https://www.vice.com/en/article/jgqeyg/facebook-is-censoring-protests-against-police-

violence-in-nigeria, accessed on 4 February 2022.

61. David Wicock ‘Instagram claims monkey emojis “DON’T breach race rules” as Twitter

removes 1,000 racist posts after appalling abuse of England stars’ Daily Mail 12 July 2021,

available at https://www.dailymail.co.uk/news/article-9780627/Racist-tweeters-knock-

door-police-face-force-law.html, accessed on 4 February 2022.

62. Emily VanDerWerff ‘#Gamergate: here’s why everybody in the video game world is

fighting’ Vox 2014, available at: https://www.vox.com/2014/9/6/6111065/gamergate-

explained-everybody-fighting, accessed on 17 February 2021.

63. Noxolo Majavu & Palesa Dlamini ‘Films and Publications Amendment Act leaves online

content producers hot under the collar’ News24 5 March 2022, available at

https://www.news24.com/citypress/news/films-and-publications-amendment-act-leaves-

producers-hot-under-the-collar-20220305, accessed on 10 October 2022.

64. Sophie Marineau, ‘Fact check US: what is the impact of Russian interference in the US

presidential election?’ The Conversation 29 September 2020, available at:

https://theconversation.com/fact-check-us-what-is-the-impact-of-russian-interference-in-

the-us-presidential-election-146711, accessed on 26 November 2020.

65. Taylor Hatmaker ‘Russia Says it is restricting access to Facebook in the country’

Techcrunch 25 February 2022 available at https://techcrunch.com/2022/02/25/russia-

facebook-restricted-censorship-ukraine/, accessed on 26 June 2022.

66. Victoria Elms & Kieran Devine ‘Euro 2020: Why is it so difficult to track down racist

trolls and remove hateful messages on social media?’ Sky News 21 July 2021, available at

https://news.sky.com/story/euro-2020-why-is-it-so-difficult-to-track-down-racist-trolls-

and-remove-hateful-messages-on-social-media-12358392, accessed on 4 February 2022.

https://www.vice.com/en/article/jgqeyg/facebook-is-censoring-protests-against-police-violence-in-nigeria
https://www.vice.com/en/article/jgqeyg/facebook-is-censoring-protests-against-police-violence-in-nigeria
https://www.dailymail.co.uk/news/article-9780627/Racist-tweeters-knock-door-police-face-force-law.html
https://www.dailymail.co.uk/news/article-9780627/Racist-tweeters-knock-door-police-face-force-law.html
https://www.vox.com/2014/9/6/6111065/gamergate-explained-everybody-fighting
https://www.vox.com/2014/9/6/6111065/gamergate-explained-everybody-fighting
https://www.news24.com/citypress/news/films-and-publications-amendment-act-leaves-producers-hot-under-the-collar-20220305
https://www.news24.com/citypress/news/films-and-publications-amendment-act-leaves-producers-hot-under-the-collar-20220305
https://theconversation.com/fact-check-us-what-is-the-impact-of-russian-interference-in-the-us-presidential-election-146711
https://theconversation.com/fact-check-us-what-is-the-impact-of-russian-interference-in-the-us-presidential-election-146711
https://techcrunch.com/2022/02/25/russia-facebook-restricted-censorship-ukraine/
https://techcrunch.com/2022/02/25/russia-facebook-restricted-censorship-ukraine/
https://news.sky.com/story/euro-2020-why-is-it-so-difficult-to-track-down-racist-trolls-and-remove-hateful-messages-on-social-media-12358392
https://news.sky.com/story/euro-2020-why-is-it-so-difficult-to-track-down-racist-trolls-and-remove-hateful-messages-on-social-media-12358392

I. INTRODUCTION
II. THE UNREALISED PROMISE OF THE INTERNET AND WHY WE MODERATE
(a)The need for content moderation
(b) The rationale for moderation and how it is defined
(c) The operation of content moderation
(d) Considerations for content moderation

III. THE RIGHT TO FREEDOM OF EXPRESSION IN SOUTH AFRICA
(a) Un-packing the right to freedom of expression in South Africa
(b) The operation and limitations of the right to freedom of expression
(c) Freedom of expression online

IV. HOW CONTENT MODERATION UNDERMINES THE RIGHT TO FREEDOM OF EXPRESSION IN SOUTH AFRICA
(a) The limitations extend beyond section 16(2)
(b) The limitations are vague
(c) Prior restraint

V. THE ROLE OF THE STATE
(a) Why involve the state?
(b) Reconceptualising the right in South Africa

VI. SOUTH AFRICA’S REGULATION OF ONLINE EXPRESSION
(a) Conditional limitation of liability
(b) The take-down notification process
(i) Incentivising removal
(ii) Outsourcing rights-based determinations

VII. CONCLUSION