The South African AI Regulatory Landscape: Reflections on Normative Asymmetry

1 Introduction 

Artificial intelligence, and generative AI in particular, represents one of the fastest-developing and most widely diffused technological shifts in modern history, surpassing even the rate of  adoption associated with the internet and smartphone technologies. The speed of AI  development, coupled with its extensive social and economic implications, has generated a  sense of heightened urgency around the creation of effective governance frameworks.1 This  urgency is not confined to the international sphere: all jurisdictions must contend with the  regulatory demands of a changing socio-technological landscape, in ways that balance  economic stability, innovation, safety and social protection.2 

South Africa is no exception.3In response to these pressures, the state launched the South  Africa National Artificial Intelligence Policy Framework (hereafter, “National Policy  Framework”) as a first step towards a comprehensive approach to AI governance;4this  trajectory has been reinforced by South Africa’s participation in the Global AI Summit for  Africa and its subsequent endorsement of the Africa Declaration on Artificial Intelligence (“the Declaration”) in April 2025, decisively signalling its readiness to play an active role in  promoting human-centred, development-oriented AI governance aligned with Global South  sensitivities and priorities.5 Despite these policy initiatives, South Africa has yet to enact AI specific legislation, leaving domestic governance to operate through a regulatory lattice of  policy instruments, existing legal frameworks, and external normative influences. 6 

In the absence of binding AI-focused regulation, AI governance in South Africa is shaped by a  complex normative landscape.7 European standards exert significant influence through  economic leverage, regulatory maturity, and institutional dissemination. By contrast, African  regional instruments articulate an open, inclusive and cooperative development philosophy  which lacks comparable mechanisms of diffusion and enforcement. The normative pluralism  created by these influences frames the conditions under which regulatory models are evaluated,  incorporated and prioritised within South Africa’s emerging AI governance framework. 

This article examines how these forces interact within this policy-driven governance landscape pending the potential promulgation of a national AI Act.8It first situates South Africa within  the broader AI policy context, highlighting influential external mechanisms that direct its current positioning. It then indicates how they are internalised by analysing the role and  authority of these instruments within South Africa’s legal and policy landscape. Comparatively  marking their reflection in the National Artificial Intelligence Policy Framework, the article  explores how urgency, regulatory credibility, and uneven normative influence shape domestic  AI governance choices prior to formal enactment. 

2 External Normative Influences on South African AI Regulation

2.1 European Union: Conditional Models and Risk-based AI Governance 2.1.1 EU AI Act Charged with the same call to proactively set appropriate governance measures in place,9the  European Commission proposed and developed a synchronous, formal mechanism to regulate  Artificial Intelligence within its region.10 In 2023, the European Parliament adopted its  position, and in May of 2024, the European Council officially adopted Regulation (EU)  The South African AI Regulatory Landscape: Reflections on Normative Asymmetry 2024/1689.11 Commonly referred to as the European Union Artificial Intelligence Act (“EU AI  Act”/ “the Act”), the Act introduces a risk-based system of regulation predominantly aimed at  developers of AI systems to ensure the protection of fundamental rights and industry alignment  with EU values.12 Depending on the level of risk a system or model is evaluated at, it may be  classified as unacceptable (prohibited) risk,13 high risk or limited to minimal risk.14 Each  classification corresponds to an appropriate degree of regulatory compliance,15 based on the  technology’s capability to deceive or manipulate, facilitate social control16, invade privacy,17 allow misuse or discriminatory behaviour, and cause social disruption.18 

The Act is directly applicable to all European Union (EU) member states as domestic law,  creating a uniform legal standard across Europe which imposes duties on each state to  implement national regulatory measures to ensure compliance.19 Jurisdictions outside the EU must comply with the legal obligations imposed by the Act insofar as a state or its national (a  natural or legal person) qualifies as a deployer or a provider as defined by the Act.20 This is  also true where the AI system or model it uses is marketed, made available, or produces output  used within the EU.21 

2.1.2 EU General Data Protection Regulation 

In 2016, the EU adopted the General Data Protection Regulation (“GDPR”),22 which enshrines  the right of natural persons to the protection of personal data.23 It similarly intended to produce  a harmonised legal framework across the EU with domestically enforceable legal regimes.24 The GDPR, like the EU AI Act, stipulates that the regulation applies to controllers and  Michaele Kemp processors or their operations within the Union even if processing occurs elsewhere;25 it applies  to the data of data subjects within the EU, insofar as goods or services are marketed to them or  their behaviour within the territory is monitored.26 The regulation is characterised by its  conditional transference of personal data dependent on compliance, data subject consent and  satisfactory national data protection laws,27 which has been attributed to its fundamentally  rights-based approach.28 

2.2 African Union: Open Models and Development-Oriented AI Governance 2.2.1 Africa Declaration on Artificial Intelligence 

The Africa Declaration on Artificial Intelligence,29 to which South Africa is a signatory, was  adopted by 52 member states of the African Union (AU) in April 2025, expressing formal  acceptance of the African Union Continental AI Strategy.30 The Declaration denotes inclusivity  and sovereignty as guiding principles which allow African values and contexts to inform and  empower AI development on the continent.31 It centres human freedom and dignity in its aim  to protect privacy and transparency, while urging development to uplift African societies and  economies.32 

Similar to the EU, the AU supports regionally coordinated national policy and regulatory  frameworks,33 and deliberately includes regulations for cross-border data transfers, 34 mirroring  the EU’s GDPR objectives. In contrast to the EU’s conditional regulatory model,35 however,  the Declaration signals the adoption of an open regulatory model through its commitment to  establishing an African network of open AI models and open data sets.36 Rather than requiring recipient countries to comply with certain codes of conduct or to satisfy a standard of national  regulation for lawful trade,37 it relies heavily on policy alignment, voluntary cooperation  The South African AI Regulatory Landscape: Reflections on Normative Asymmetry  between its members, and self-sworn commitment to abide by broader governance  frameworks.38 

2.2.2 African Union Data Policy Framework 

In February 2022, the African Union Executive Council formally endorsed the African Union  Data Policy Framework (AUDPF)39 to aid its member states in creating domestic data  regulations informed by African development-centred objectives and guiding principles.40 The  greater purpose thereof is to economically empower African nations while strengthening the  continent’s global position in the digital data-driven economy of the modern era.41 Though it,  too, supports the individual right of data subjects in claiming protection of personal data,42 its  approach situates data subjects within the broader “data economy,”43 calling for a balance  between stringent protection regulation and equally risk-mitigating and data flow-supportive  policy.44 As it has no binding power, the policy emphasises the need for national legal  frameworks to be revised in accordance with the policy tenets, taking into account their  respective domestic contexts.45 However, it also recognises that a harmonised continental legal  approach would minimise uncertainty, bolster public trust and enable fair trade and  competition.46 Like the Africa Declaration, it endorses open data flow to foster inclusive access  to the data economy,47 subject to the same principles of transparency, accountability, consent  and legitimacy (lawfulness),48 and promotes cooperative governance efforts to boost industry  capabilities equitably while upholding human rights.49 Norm Transposition and Regulatory Asymmetry: The “Brussels Effect” The EU’s ability to transpose European norms and standards in the digital space is often  attributed to its market size and economic strength.50 Yet its regulatory leadership is equally  shaped by the speed with which it has acted as a first-mover in digital governance on the global stage.51 EU legislative acts exert a deeply integrative regional effect, not through classical  monism, but through an autonomous and constitutionalised legal order in which regulations  are directly applicable and enjoy primacy over national law.52 This unified and democratically grounded governance framework stabilises the internal legal order and confers external  legitimacy.53 

The combination of economic power and harmonised internal politics incentivises the  extraterritorial adoption of its conditional regulatory model.54 Such adoption occurs either through voluntary compliance (de facto), largely for access and ease of trade, or through formal  adoption of the legislative or normative template by governments outside the EU (de jure).55 Many EU norms and standards have permeated the Union’s territorial borders in this way,  constituting an observable trend termed the “Brussels effect.”56 The EU’s General Data  Protection Regulation is an instance thereof. In the years following the GDPR’s  implementation, many African countries followed legislative suit, strongly leaning on the  GDPR in formulating domestic data protection laws.57 

South Africa has experienced this dynamic, too. A legislative act similar to the GDPR was  promulgated in South Africa as early as 2013, but the Protection of Personal Information Act (POPIA) only came into full effect in 2020.58 POPIA was enacted to honour the constitutionally  enshrined right to privacy, which extends to the protection of one’s private communications [or  data] from infringement.59 It bears a notable resemblance to,60 and in reference acknowledges,61 the GDPR in terms of certain definitions,62 data subject consent on data  transfers,63 principles of accountability and transparency,64 and attributable rights.65 The EU’s  conditional regulatory model is reflected in the conditions POPIA sets out in section 4(1) for  The South African AI Regulatory Landscape: Reflections on Normative Asymmetry the lawful processing of personal data to which the respective data subject has a legal right.66 Section 72 likewise adopted language akin to the GDPR in setting conditions to transfer personal information across borders, requiring third-party recipients to have obtained the  consent of the data subject67 and possess some form of satisfactory regulation that adheres to  the same principles of lawfulness.68 As AI is a data-driven innovation, POPIA will be relied on extensively to govern many aspects of AI in the present legal framework.69 The normative position of the GDPR, particularly on data governance, is concordant with that  of the EU AI Act.70 The regulations operate in conjunction to give effect to a harmonised legal  framework, sensitive to threats against individual rights to data protection and broader societal  disruption.71 By virtue of its GDPR-integrated data protection legislation, South Africa is  predisposed to embedding a European regulatory framework as a model for developing its  domestic AI governance. Moreover, given South Africa’s significant trade relationships with  the European Union, the Brussels Effect is likely to manifest in a de facto manner in the  interim,72 and potentially in de jure form should policy alignment give rise to legislative  action.73 

4 African Commitments and Structural Constraints  

This predisposition towards European regulatory models, however, does not exhaust the  normative landscape within which South Africa’s AI governance is being shaped. At the  regional level, South Africa has also endorsed African normative instruments,74 including the  Africa Declaration on Artificial Intelligence and the AU Data Policy Framework, which  articulate a distinct approach to data and AI governance. These instruments emphasise  development, inclusion,75 and the creation of enabling data ecosystems,76 reflecting regulatory  The South African AI Regulatory Landscape: Reflections on Normative Asymmetry 5 Governance Tensions in South Africa’s National AI Policy Framework The South African National AI Policy Framework reflects the tensions arising from South  Africa’s position within multiple normative orders, while simultaneously articulating a  transformative vision for AI innovation.85 In line with its African regional commitments, the  Framework advances a contextual and sector-specific approach to AI governance,86 emphasising developmental responsiveness,87 institutional capacity,88 and the use of data as a  resource for social and economic inclusion.89 In this respect, it mirrors the enabling and  cooperative regulatory language found in the AUDPF and the Africa Declaration, 90 particularly  in its emphasis on open but safe access to data. 91 

At the same time, it prioritises legal reform aimed at strengthening data protection and  regulatory oversight, including through existing legislation such as the GDPR-influenced  POPIA.92 This dual emphasis gives rise to uncertainty as to whether South Africa’s AI  governance trajectory will continue to draw more heavily on European-style conditional  regulatory models,93 particularly in relation to risk mitigation and compliance structures. 94 This  uncertainty is reflected in consultative materials underlying the National AI Policy Summit,95 which acknowledge both the regulatory sophistication and democratic legitimacy of the EU AI  Act, as well as the urgency created by the rapid pace of AI development.96 Under such  conditions, the reliance on readily available and credible external governance models may  shape domestic regulatory choices,97 even where these sit in tension with longer-term  commitments to open, development-oriented AI governance articulated at the regional level.98 priorities that differ in important respects from the conditional, risk-based model underpinning  European AI governance.77 

While both responses acknowledge individual rights, the AU decentres it as a priority, framing  AI as a tool for social and developmental upliftment.78 it endorses an open regulatory model  that is conducive to intergovernmental collaboration and institutional cooperation to overcome  foundational disparities.79 This emphasis on cooperative governance functionally addresses  Africa’s weaker institutional standing,80 which would make a compliance-heavy framework  economically burdensome and render AI resources inaccessible. Neither the Africa Declaration  nor the AUDPF are legislative, nor do they have binding force; South Africa, as a dualist system  of international law incorporation, requires implementation through formal incorporation into  national legislation.81 These frameworks still bear significant interpretive and, thereby,  normative value as soft law instruments, however, offering imperative institutional direction. 

The absence of binding force and enforcement mechanisms means that these African  frameworks rely on deliberate political and legislative uptake, as opposed to structural or  market-driven compulsion.82 In contrast to EU regulatory models, which acquire traction  through economic leverage and immediate compliance incentives, African instruments require  conscious normative prioritisation to shape domestic governance outcomes. Even if the AU  were to produce an AI Act which South Africa could implement, it too would be subject to a  process of parliamentary approval.83 These modes of policy distillation, therefore, entail slower  processes of adoption and legislative reform, which are less well suited to the rapid pace of AI  development.84 

Though South Africa is well-equipped to legislate operable and cohesive AI legislation, it  should remain cognisant of this normative asymmetry. Unreflective reliance on dominant  external models during the transition from policy to legislation, particularly when adopted  under conditions of urgency, may risk entrenching approaches that perpetuate legal uncertainty. 

6 Conclusion 

The global acceleration of artificial intelligence development has intensified the demand for  timely and credible governance frameworks capable of balancing innovation, economic  interest and maintaining social order. However, for states like South Africa, the urgency driving  AI regulation risks foreclosing the opportunity to make intentional, context-sensitive, and long term philosophically aligned legal decisions. As this article has shown, the Brussels effect  illustrates how the availability of comprehensive and enforceable external regulation,  supported by institutional capacity and market leverage, may shape domestic regulatory choice,  where comparable regional frameworks lack equivalent legislative voice and internal cohesion. 

The European Union Artificial Intelligence Act represents a sophisticated, democratically legitimised approach to AI governance, grounded in a human-centric focus on protecting  individual rights, digital sovereignty and regional innovation. In this respect, the Act is not  misaligned with South Africa’s emerging AI governance vision, as articulated in the National  Artificial Intelligence Policy Framework, and reflected in the Africa Declaration on Artificial  Intelligence. Nevertheless, a tension emerges where the conditional, risk-based regulatory logic  underpinning European AI governance interacts with African regional commitments that  require strategically open and cooperative capacity-building efforts. 

While the adoption of readily available, comprehensive and credible external regulatory  models may be understandable under conditions of urgency, the entrenchment of such models  carries implications for South Africa’s longer-term commitments to technological sovereignty,  contextual responsiveness, and regional development. Although a comprehensive assessment  of alternative regulatory pathways lies beyond the scope of this article, the analysis underscores  the importance of conscious legislative choice in shaping AI governance regimes that do not  merely reflect dominant external norms but are aligned with South Africa’s constitutional  values and developmental priorities.

The South African AI Regulatory Landscape: Reflections on Normative Asymmetry 

7 Bibliography 

African Union, ‘Africa Declaration on Artificial Intelligence’ (4 April 2025), Global AI Summit  on Africa African Union, AU Data Policy Framework (February 2022), endorsed by the Executive  Council at its 40th Ordinary Session (Decision EX.CL/Dec.1144(XL)), Addis Ababa Bendiek A and Stuerzer I, ‘The Brussels Effect, European Regulatory Power and Political  Capital: Evidence for Mutually Reinforcing Internal and External Dimensions of the Brussels  Effect from the European Digital Policy Debate’ (2023) 2 Digital Society Department of Communications and Digital Technologies, South Africa National Artificial  Intelligence Policy Framework (2024) https://www.dcdt.gov.za/sa-national-ai-policy framework/file/338-sa-national-ai-policy-framework.html accessed 11 January 2026 Coetzee H and Stoch E, ‘South Africa’s AI Trajectory: Navigating the Divide Between National  Ambition and Market Reality’ (2025) 

Constitution of the Republic of South Africa,1996 

European Parliament, ‘Parliament’s Legislative Powers’ European Parliament https://www.europarl.europa.eu/about-parliament/en/parliaments-powers/legislative powers accessed 11 January 2026 

Ferracane MF, González Ugarte S and van der Marel E, ‘The Brussels Effect in Africa: Is It  Beneficial for Intra-Regional Trade in Digital Services?’ (2025) 28(1) Journal of International  Economic Law 1 

Future of Life Institute, ‘European Union Artificial Intelligence Act: Historic Timeline’  Future of Life Institute https://artificialintelligenceact.eu/developments/ accessed 11 January  2026 

Flaminio Costa v ENEL (Case 6/64) EU:C:1964:66, [1964] ECR 585 

Ka Mtuze SS, ‘Towards Drafting Artificial Intelligence (AI) Legislation in South Africa’ (trans  Morige M) (2024) 45(1) Obiter 

Regulation (EU) 2016/679 of the European Parliament and of the Council of 27 April 2016 on  the protection of natural persons with regard to the processing of personal data and on the  free movement of such data (General Data Protection Regulation) [2016] OJ L 119/1,  corrigendum [2018] OJ L 127/2 

Regulation (EU) 2024/1689 of the European Parliament and of the Council of 13 June 2024  laying down harmonised rules on artificial intelligence and amending U)EC) No 300/2008,  (EU) No 167/2013, (EU) No 168/2013, (EU) 2018/858, (EU) 2018/1139 and (EU) 2019/2144  and Directives 2014/90/EU, (EU) 2016/797 and (EU) 2020/1828 (Artificial Intelligence Act)

South African Department of Communications and Digital Technologies, ‘Discussion  Document – AI National Government Summit’ (October 2023)  https://www.dcdt.gov.za/images/phocadownload/AI_Government_Summit/National_AI_Go vernment_Summit_Discussion_Document.pdf accessed 11 January 2026

1 African Union, ‘Africa Declaration on Artificial Intelligence’ (4 April 2025), Global AI Summit on Africa,  Kigali, Rwanda; South Africa, National Artificial Intelligence Policy Framework (2024) 3

2 Department of Communications and Digital Technologies, South Africa National Artificial Intelligence Policy  Framework (2024) 4 https://www.dcdt.gov.za/sa-national-ai-policy-framework/file/338-sa-national-ai-policy framework.html accessed 11 January 2026. 

3 Heidi Coetzee and Eric Stoch, ‘South Africa’s AI Trajectory: Navigating the Divide Between National  Ambition and Market Reality’ (2025) para 2.3. The authors identify this diffusion trend within the South  African corporate sphere of rapid adoption with little to no governance strategy, risk management or oversight.

4 South Africa National Artificial Intelligence Policy Framework (n2) 3. 

5 South Africa National Artificial Intelligence Policy Framework (n2) 1,4; Coetzee and Stoch (n 2) para 4.

6 Sizwe Snail Ka Mtuze, ‘Towards Drafting Artificial Intelligence (AI) Legislation in South Africa’ (trans  Masego Morige) (2024) 45(1) Obiter 168-169.

7Ibid 167. 

8 South Africa National Artificial Intelligence Policy Framework (n2) 3. 

9 Annegret Bendiek and Inga Stuerzer, ‘The Brussels Effect, European Regulatory Power and Political Capital:  Evidence for Mutually Reinforcing Internal and External Dimensions of the Brussels Effect from the  European Digital Policy Debate’ (2023) 2 Digital Society 3. Proactive, democratized governance measures fall  within the strategic goals of the European Union to establish digital sovereignty both internally and externally. 

10 Regulation (EU) 2024/1689 of the European Parliament and of the Council of 13 June 2024 laying down  harmonised rules on artificial intelligence and amending Regulations (EC) No 300/2008, (EU) No 167/2013,  (EU) No 168/2013, (EU) 2018/858, (EU) 2018/1139 and (EU) 2019/2144 and Directives 2014/90/EU, (EU)  2016/797 and (EU) 2020/1828 (Artificial Intelligence Act) para 1; Future of Life Institute, ‘European Union  Artificial Intelligence Act: Historic Timeline’ Future of Life Institute  (https://artificialintelligenceact.eu/developments/) (accessed 11 January 2026) .

11 Future of Life Institute, Historic Timeline (n11); European Parliament, ‘Parliament’s Legislative Powers’  European Parliament (https://www.europarl.europa.eu/about-parliament/en/parliaments-powers/legislative powers) (accessed 11 January 2026). 

12 Artificial Intelligence Act (n 10), paras 8-9, 26-27. 

13 ibid, arts 5, 6. 

14 Ibid. para 53. 

15 ibid para 65; subject to continuous assessment of risk. 

16 ibid para 28. 

17 Ibid para 10. 

18 ibid para 29. 

19 ibid para 10, 153. 

20 ibid arts 2(4), 3(3)-(4) Public authorities and international organisations are exempted where the AI is used  pursuant to international agreements or cooperation, and other safeguards are put in place.

21 ibid art 2 (1)(a),(c). 

22 Regulation (EU) 2016/679 of the European Parliament and of the Council of 27 April 2016 on the protection  of natural persons with regard to the processing of personal data and on the free movement of such data  (General Data Protection Regulation) [2016] OJ L 119/1, corrigendum [2018] OJ L 127/2.

23 General Data Protection Regulation para 1, art 1(2)-(3). 

24 Martina F Ferracane, Simón González Ugarte and Erik van der Marel, ‘The Brussels Effect in Africa: Is It  Beneficial for Intra-Regional Trade in Digital Services?’ (2025) 28(1) Journal of International Economic Law 1, 5; Artificial Intelligence Act (n 10) para 10.

25 General Data Protection Regulation, art 3(1). 

26 Ibid, art 3(2)(a)-(b). 

27 Ferracane, González Ugarte and van der Marel (n 24) 5. 

28 Ibid. 

29 Africa Declaration (n 1). 

30 ibid para 1. 

31 ibid para 2. 

32 ibid para 2. 

33 ibid para 3.6.1. 

34 ibid para 3.6.3 

35 Ferracane, González Ugarte and van der Marel (n 24) 5-6 

36 Ferracane, González Ugarte and van der Marel (n 24) 5-6; Africa Declaration (n 1) para 3.2.1.

37 Ferracane, González Ugarte and van der Marel (n 24) 6; Africa Declaration (n 1) para 3.6.

38 Ferracane, González Ugarte and van der Marel (n 24) 6; Africa Declaration (n 1) para 3.6.3

39 African Union, AU Data Policy Framework (February 2022), endorsed by the Executive Council at its 40th  Ordinary Session (Decision EX.CL/Dec.1144(XL)), Addis Ababa. 

40 AU Data Policy Framework (n 39) IV, VI. 

41 ibid VI. 

42 ibid, VII. 

43 ibid VI. 

44 ibid VI. 

45 ibid VI-VIII. 

46 ibid VII-VIII, IX. 

47 ibid VIII. 

48 ibid VII. 

49 ibid VIII. 

50 Bendiek and Stuerzer (n 10) 3-4.

51 Bendiek and Stuerzer (n 10) 4; Ka Mtuze (n 6) 178. 

52 Case 6/64 Flaminio Costa v ENEL EU:C:1964:66, [1964] ECR 585. 

53 Bendiek and Stuerzer (n 10) 3-4; Artificial Intelligence Act paras 1- 3. 

54 Bendiek and Stuerzer (n 10) 2,19; Ferracane, González Ugarte and van der Marel (n 24) 10.

55 Ferracane, González Ugarte and van der Marel (n 24) 5-6. 

56 Africa Declaration (n 1). 

57 Ferracane, González Ugarte and van der Marel (n 24) 6-7. 

58 Ibid 7. 

59 Constitution of the Republic of South Africa, 1996 s14(d); Protection of Personal Information Act 4 of 2013  (POPIA), Preamble, s2(a). 

60 Ka Mtuze (n 6) 169-170. 

61 POPIA s2(2). 

62 POPIA s1; General Data Protection Regulation, art 4. 

63 Ferracane, González Ugarte and van der Marel (n 24) 7. 

64 General Data Protection Regulation, arts 5(1)(a), 5(2); POPIA s4(1)(a), (h). 

65 POPIA s2(a), (c).

66 POPIA s5; General Data Protection Regulation, arts 5,6. 

67 Ibid s72(1)(b). 

68 POPIA s72(1)(a)(i)-(ii). 

69 Ka Mtuze (n 6) 167-169. Note that the Protection of Personal Information Act is not the sole piece of  legislation governing sophisticated technologies that involve automated decision-making without human  oversight. It is, however, effective and well-enforced, making it a prudent mechanism for AI governance,  although it is not fully comprehensive in scope of the National AI Policy Framework.

70 Artificial Intelligence Act, art 10. 

71 Artificial Intelligence Act paras, 7-8,10,21,26-27,67,70. 

72 South African Department of Communications and Digital Technologies, ‘Discussion Document – AI  National Government Summit’ (October 2023) 41  https://www.dcdt.gov.za/images/phocadownload/AI_Government_Summit/National_AI_Government_Summit _Discussion_Document.pdf accessed 11 January 2026. 

73 Ferracane, González Ugarte and van der Marel (n 24) 7-8. 

74 Ka Mtuze (n 6) 171-174. This source is included to demonstrate the strides the African Union, and South  Africa in particular, have made in way of crafting an Africa-empowering AI policy framework. 75 South Africa National Artificial Intelligence Policy Framework (n2) 10, para 5.4.2. 76 AU Data Policy Framework (n 39) VI.

77 Coetzee and Stoch (n 2) paras 1.1, 1.2. 

78 Africa Declaration (n 1) para 2.2.1. 

79 Ibid 3.2.1. 

80 AU Data Policy Framework (n 39) VI. 

81 Constitution of the Republic of South Africa,1996 s231 provides for the incorporation of international  agreements into domestic law through an Act of Parliament or by approval of the National Assembly and the  National Council of Provinces, subject to certain exceptions. 

82 Coetzee and Stoch 2025, [2.3] The data produced on the adoption of AI technologies within South African  enterprises shows that the business sector, inter alia, occurs largely without thorough governance  frameworks or strategies. The urgency created by market pressure demonstrably incentivises actors to pursue  technological adaptation for productivity gains over implementing ethically and purpose- aligned operational  governance. 

83 Constitution of South Africa,1996, s231. 

84 South Africa National Artificial Intelligence Policy Framework (n2) 6-7.

85 South Africa National Artificial Intelligence Policy Framework (n2) 3,6. 

86 South Africa National Artificial Intelligence Policy Framework (n2) 3,6; Coetzee & Stoch 2025, The data and  insight gathered by the authors demonstrate the practicality and futurity of this policy approach. 87 Africa Declaration (n 1) para 3.4. 

88 South Africa National Artificial Intelligence Policy Framework (n2) 1; Africa Declaration (n 1) paras 3.3, 3.4.

89 South Africa National Artificial Intelligence Policy Framework (n2) 3. 

90 South Africa National Artificial Intelligence Policy Framework (n2) 10, para 6.6 ; Africa Declaration (n 1) para 3.2. 

91 South Africa National Artificial Intelligence Policy Framework (n2) 10, paras 6.6, 6.9; Africa Declaration (n1) para 3.2.1. 

92 South Africa National Artificial Intelligence Policy Framework (n2) 10, para 6.6. 

93 ‘AI National Government Summit Discussion Document’ (n 72) 41. 

94 South Africa National Artificial Intelligence Policy Framework (n2) 10, para 6.7; Coetzee and Stoch 2025,  The lack of sectoral or corporate governance regarding AI suggests that South African regulation might  benefit from a risk-informed orientation. 

95 ‘AI National Government Summit Discussion Document’ (n 72). 

96 ‘AI National Government Summit Discussion Document’ (n 72) 40-41. 

97 Ka Mtuze (n 6) 168. 

98 Ibid 178 -179. The authors urge South Africa to legislate, as “not just a South African priority but an African  priority.”