Authored By: Amahle Mfana
University of Johannesburg
Introduction
One of the most revolutionary technologies of the twenty-first century, artificial intelligence (AI) is changing government systems, economics, and societies all over the world.[1] AI has developed into a potent instrument that can carry out activities that have historically required human intelligence, from its early beginnings in expert systems of the 1980s to the emergence of machine learning and deep learning in the last 20 years. These consist of autonomous decision-making, facial recognition, natural language processing, and predictive analytics.[2] Governments must reevaluate how laws and policies should react to technology disruption because of the potential and hazards brought about by the rapid rate of development.[3]
Jurisdictions around the world have taken varying stances on regulating AI.[4] With the AI Act (2024), a thorough risk-based framework that classifies AI systems based on their potential harm, the European Union has taken the lead.[5] The US has chosen a sectoral strategy, depending on organizations like the Federal Trade Commission to publish rules about consumer protection and fairness.[6] As part of its larger focus on government control over digital technologies, China has imposed stringent regulations on algorithmic recommendation systems.[7] International organizations like the OECD and UNESCO have also released ethical guidelines and principles, highlighting the necessity of international collaboration in AI governance.[8]
In contrast, South Africa has not yet passed any laws pertaining to AI.[9] Rather, AI is indirectly regulated through the Constitution of the Republic of South Africa’s, 1996, the Protection of Personal Information Act 4 of 2013 (POPIA), the Promotion of Access to Information Act 2 of 2000 (PAIA), the Electronic Communications and Transactions Act 25 of 2002 (ECTA), the Cybercrimes Act 19 of 2020, and the Promotion of Administrative Justice Act 3 of 2000 (PAJA).[10] Since these laws were enacted before to the development of contemporary AI technology, they do not address concerns like algorithmic bias, explainability, or responsibility for autonomous systems, even if they offer significant protections for privacy, equality, dignity, and administrative justice.[11]
Due to the nation’s constitutional commitment to human rights, the South African setting is especially important.[12] The Bill of Rights, found in Chapter 2 of the Constitution, upholds principles including equality, privacy, freedom of speech, and dignity.[13] AI technologies directly affect these rights.[14] For instance, automated content creation may undermine freedom of expression, facial recognition systems may violate privacy, and predictive policing may threaten equality.[15] As a result, the constitutional framework offers both a basis and a challenge: although it gives guidelines for regulating AI, it also requires that laws be strong enough to safeguard rights in the digital era.[16]
AI also offers South Africa significant economic prospects.[17] AI is being used in the financial industry by banks like Tyme Bank and insurers like Discovery to enhance risk assessment and customer service.[18] AI is utilized in agriculture to improve sustainability and productivity through precision farming. Predictive maintenance systems lower expenses and increase safety in mining.[19] These uses show how AI may spur development and innovation.[20] However, they also run the risk of increasing algorithmic bias, umfairness, and a lack of accountability in the absence of explicit regulations.[21]
Therefore, this article questions whether comprehensive legislation is necessary or if South Africa’s current legal framework effectively governs AI.[22] It contends that although current legislation offers significant protections, it is nevertheless disjointed and unprepared to handle the difficulties presented by AI.[23] This article concludes that South Africa should implement a dedicated AI regulatory framework that strikes a balance between innovation and constitutional rights after analysing constitutional case law, critically assessing existing legislation, comparing South Africa’s strategy with the European Union AI Act.[24]
Analysis of Case Law:
The Constitutional Court and other higher courts have created principles that are extremely pertinent to AI legislation, even if South Africa has not yet produced case law directly addressing AI.[25] These values pertain to administrative justice, accountability, equality, privacy, and dignity. [26]
Dignity and Privacy:
In NM v. Smith 2007 (5) SA 250 (CC), the Court determined that the applicants’ fundamental rights to privacy and dignity were breached by the unapproved publishing of private medical information. The Court acknowledged that people have a right to assume that their private information will be kept private. AI systems, which frequently handle enormous amounts of personal data, can directly benefit from this logic. This idea is strengthened by POPIA, which mandates that responsible parties handle personal data in a fair and legal manner.
Equality and Freedom of Expression
The Court emphasized in Khumalo v Holomisa 2002 (5) SA 401 (CC), that the common law must be created in accordance with constitutional ideals, such as equality, freedom of expression, and dignity. Because recommendation algorithms and automated content creation are rapidly influencing public conversation, this idea is important for AI. To guarantee that AI-driven communication upholds constitutional rights, courts may need to modify current doctrines.[27]
Equality in Substance:
The Court emphasized the significance of substantive equality and embracing diversity in MEC for Education: KwaZulu-Natal v. Pillay 2008 (1) SA 474 (CC). AI systems that have been trained on biased datasets may inadvertently discriminate against people based on socioeconomic class, gender, color, or handicap. Pillay shows that constitutional guarantees of equality cannot be compromised by computerized decision-making.
Justice in Administration:
When government agencies deploy AI, administrative law principles are equally pertinent. Administrative decisions must be rational, legal, and procedurally fair, according to PAJA. Affected parties must have the right to know how judgments were made and to contest unfair results if public authorities use AI to make decisions about social grants, immigration, or law enforcement.
When taken as a whole, these instances offer a constitutional basis for regulating AI. They do not, however, address issues unique to AI, such as algorithmic transparency, autonomous system liability, or high-risk application regulation.
Critical Assessment:
A disjointed structure results from South Africa’s reliance on general statutes. Although PAJA guarantees justice in administrative decisions and POPIA safeguards privacy, these rules were passed prior to the widespread use of AI. They don’t deal with explainability, algorithmic bias, or culpability. Researchers contend that this leads to ambiguity for regulators and developers (Naidoo, 2024).
Liability for damage caused by autonomous systems is still uncertain due to the lack of AI-specific legislation. Although common law concepts like vicarious liability and negligence may be applicable, they are not well-suited to handle complicated accountability issues when judgments are made by algorithms instead of people.
Moreover, there are few obligations for transparency. POPIA does not require the disclosure of algorithmic reasoning, but it does require responsible parties to notify data subjects about the gathering of personal information. This makes it more difficult for people to contest unjust automatic decisions.
Scholarly analysis highlights how South Africa’s fragmented governance erodes public confidence. According to Phatsoane (2024), the introduction of AI runs the potential of undermining constitutional rights in the absence of clear regulations.According to Chilunjika and Uwizeyimana (2025), African states frequently implement technology without sufficient governance frameworks, which results in ambiguity and uneven enforcement.
Comparative Evaluation: EU AI Act:
A helpful model is offered by the EU AI Act (2024). It divides AI systems into four risk categories: • Unacceptable risk: It is forbidden for AI systems to exploit social score or manipulate behaviour. • High risk: Strict requirements, such as paperwork, human monitoring, and explainability, apply to AI employed in healthcare, finance, law enforcement, and employment. • Minimal risk: Chatbots and other systems must reveal that users are engaging with AI. • Low risk: Applications like spam filters are still mostly uncontrolled. This risk-based strategy strikes a balance between human rights protections and innovation. According to Veale and Borgesius (2021), it guarantees proportionality by enforcing more stringent duties exclusively in areas with the highest risks.
There is no such classification in South Africa. Regardless of the degree of risk, all AI systems are implicitly governed by generic statutes. This leads to uncertainty and inefficiency. Additionally, by mandating that member states appoint national supervisory authorities, the EU AI Act creates institutional control. There is no comparable organization in South Africa.
Suggestions:
First, detailed AI-specific laws need be passed in South Africa. AI-specific issues are not addressed by current legislation like POPIA, PAJA, ECTA, and PAJA, which offer indirect control. Consistent standards and legal certainty would be provided by dedicated legislation (Naidoo, 2024). Second, like the EU AI Act (2024), legislation should take a risk-based approach. Minimal-risk systems should continue to be loosely controlled, while high-risk applications should be subject to stringent requirements (Veale and Borgesius, 2021).
Thirdly, accountability and openness ought to be required. Businesses using AI should explain automated judgments and document how systems function (Phatsoane, 2024).
Fourth, South Africa ought to create a separate regulatory body with AI expertise. AI poses more general concerns like ethics and prejudice, even though the Information Regulator oversees data protection (Chilunjika and Uwizeyimana, 2025). Ultimately, funding AI research and teaching is crucial. To advance digital literacy and ethical AI, universities and private stakeholders should work together (Mlondolozi and Masumbe, 2026).
Conclusion:
The business and society of South Africa could be completely changed by artificial intelligence. However, laws created prior to the creation of modern AI technology are unable to adequately handle the complex issues raised by its quick development.
According to this article, AI is only indirectly regulated by the Constitution, POPIA, PAJA, and ECTA, even though they offer significant protections. The benefits of specialized legislation that takes a risk-based approach and offers efficient oversight are demonstrated by the comparative analysis of the EU AI Act (2024). As a result, South Africa ought to embark on creating AI-specific laws to supplement its current system. South Africa can promote responsible innovation, boost public trust in AI, and guarantee that technological advancement is in line with constitutional rights and the public interest by implementing a thorough regulatory framework.
Bibliography:
1 Legislation
- Constitution of the Republic of South Africa, 1996.
- Protection of Personal Information Act 4 of 2013.
- Promotion of Access to Information Act 2 of 2000.
- Electronic Communications and Transactions Act 25 of 2002.
- Cybercrimes Act 19 of 2020.
- Promotion of Administrative Justice Act 3 of 2000.
2 Cases
- NM v Smith 2007 (5) SA 250 (CC).
- Khumalo v Holomisa 2002 (5) SA 401 (CC).
- MEC for Education: KwaZulu-Natal v Pillay 2008 (1) SA 474 (CC).
3 Secondary Sources
- Russell S and Norvig P, Artificial Intelligence: A Modern Approach (4th edn, Pearson 2021).
- Naidoo, ‘Artificial Intelligence Regulation in South Africa: Bridging the Gap’ (2024) South African Journal of Law and Technology.
- Veale M and Borgesius FZ, ‘Demystifying the EU’s AI Act’ (2021) Computer Law & Security Review.
- Phatsoane, ‘AI Governance and Legal Certainty in South Africa’ (2024) UJ Law Review.
- Chilunjika and Uwizeyimana, ‘Artificial Intelligence and Governance in Africa’ (2025) African Journal of Public Affairs.
- Mlondolozi and Masumbe, ‘Epistemological Challenges of AI in Research and Regulation’ (2026) South African Journal of Higher Education.
- OECD, OECD Principles on Artificial Intelligence (2019).
- UNESCO, Recommendation on the Ethics of Artificial Intelligence (2021).
[1] Stuart Russell and Peter Norvig, Artificial Intelligence: A Modern Approach (4th edn, Pearson 2021).
[2] Ryan Abbott, The Reasonable Robot: Artificial Intelligence and the Law (Cambridge University Press 2020).
[3] UNESCO, Recommendation on the Ethics of Artificial Intelligence (UNESCO 2021).
[4] OECD, OECD AI Principles (OECD 2019).
[5] Regulation (EU) 2024/1689 of the European Parliament and of the Council laying down harmonised rules on artificial intelligence (Artificial Intelligence Act).
[6] Federal Trade Commission, Using Artificial Intelligence and Algorithms (FTC Guidance).
[7] Interim Measures for the Management of Generative Artificial Intelligence Services (China, 2023).
[8] UNESCO, Recommendation on the Ethics of Artificial Intelligence (UNESCO 2021); OECD, OECD AI Principles (OECD 2019).
[9] SSK Mtuze and M Morige, ‘Towards Drafting Artificial Intelligence (AI) Legislation in South Africa’ (2024) Obiter.
[10] Constitution of the Republic of South Africa, 1996; Protection of Personal Information Act 4 of 2013; Promotion of Access to Information Act 2 of 2000; Electronic Communications and Transactions Act 25 of 2002; Cybercrimes Act 19 of 2020; Promotion of Administrative Justice Act 3 of 2000.
[11] V Mbonye, M Moodley and F Nyika, ‘Examining the Applicability of the Protection of Personal Information Act in AI-Driven Environments’ (2024) South African Journal of Information Management.
[12] Constitution of the Republic of South Africa, 1996.
[13] Constitution of the Republic of South Africa, 1996 ch 2.
[14] Mtuze and Morige (n 9).
[15]UNESCO (n 8).
[16] Constitution of the Republic of South Africa, 1996 ss 9, 10, 14 and 16.
[17] OECD (n 4).
[18] Mtuze and Morige (n 9).
[19] OECD (n 4).
[20] Ryan Abbott (n 2).
[21] UNESCO (n 8).
[22] Mtuze and Morige (n 9).
[23] Mbonye, Moodley and Nyika (n 11).
[24] Regulation (EU) 2024/1689 (Artificial Intelligence Act).
[25] Constitution of the Republic of South Africa, 1996.
[26] Constitution of the Republic of South Africa, 1996 ch 2.
[27] NM and Others v Smith and Others 2007 (5) SA 250 (CC).
