Authored By: Lindelwa Lerato Mlaba
University of South Africa
Introduction
As artificial intelligence has become deeply embedded in society, exerting a huge influence on decision-making processes across public sectors such as healthcare, education, law enforcement and also the private sector.[1] South African law must address both the opportunities and regulatory challenges posed by its introduction. AI is transforming how both public and private organisation’s function. These technological developments have also introduced legal and ethical challenges that the country’s existing legal frameworks were not specifically designed to address.[2]
These challenges have also compelled our government to reconsider whether the current legal frameworks are adequate for regulating AI technologies while safeguarding fundamental rights. While South Africa has enacted partial regulations such as the 1996 Constitution, the Protection of Personal Information Act 4 of 2013(POPIA), the Cybercrimes Act 19 of 2020, and the Electronic Communications and Transactions Act 25 of 2002, the country currently lacks a comprehensive legal framework specifically governs artificial intelligence. [3]
This article argues that although South Africa has a constitutional framework providing a solid foundation for responsible AI use, the lack of specific legislation creates uncertainty that stifies innovation and exposes vulnerable populations to risks. This article further contends that South Africa should adopt a new, comprehensive risk-based regulatory framework for Artificial Intelligence that fosters innovation while safeguarding the constitutional rights of its citizens, specifically those outlined in sections 9, 10, 14 and 34 of the Constitution.[4] The Constitutional Court has consistently affirmed these rights as fundamental values of South Africa’s constitutional democracy, as evidenced by the National Artificial Intelligence Policy Framework (2024), UNESCO’s recommendations on the Ethics of Artificial Intelligence (2021), the Protection of Personal Information Act (2013), the Electronic Communications and Transactions Act (2002), the Cybercrimes Act (2020), and the Constitution of the Republic of South Africa (1996), particularly in the light of S v Makwanyane 1995 (3) SA 391 (CC) para 144.[5]
Defining the concept of Artificial Intelligence
Artificial intelligence refers to computer systems designed to perform tasks that typically require human instincts or intelligence. These tasks encompass reading data, solving complex problems, accessing to more additional information for specific searches, and making decisions. AI systems function as modernised machines that deliver algorithms and provide comprehensive answers and solutions to any problem. Generative AI systems have emerged as powerful tools capable of producing content, written documents, possible questions and answers, images, videos and software codes. In my opinion, AI systems have already taken over the world, representing the fastest rising innovation to expand so rapidly in such a short period.
Governments and businesses have demonstrated a reliance on AI to boost efficiency, cut operational costs and improve decision-making. In South Africa, AI technologies are employed across various sectors including banking for fraud detection, educational for personalised learning and law enforcement for crime surveillance. AI also poses legal and ethical challenges due to its often-automated decision-making processes. This “black box” phenomenon, complicates the assignment of legal liability when AI systems generate harmful and discriminatory outcomes that endanger the public and other systems. Consequently, questions emerge about accountability when AI systems cause financial loss or make unlawful decisions, given that as mere tools executing tasks based on requests, they cannot be held responsible for the harm they inflict. These unanswered questions highlight why AI has emerged as a defining legal challenge of the Fourth Industrial Revolution. AI has profoundly impacted decisions affecting individual’s rights, necessitating that the law ensure technological advancements align with constitutional values and the rule of law.[6]
Current Legal Framework Governing Artificial Intelligence in South Africa
Although South Africa currently lacks specific legislation that governing artificial intelligence, relevant constitutional provisions and statutes have been enacted to regulate various aspects of AI, as reflected in cases such as S v Makwanyane 1995 (3) SA 391 (CC) para 144 and Harksen v Lane NO 1998 (1) SA 300 (CC). These laws have established key legal principles concerning privacy, equality, accountability, cybercrime, consumer protection and administrative justice.[7] The next section will discuss these provisions in detail.
I)The Constitution of the Republic of South Africa, 1996
In South Africa, the 1996 Constitution serves as the supreme law and establishes the primary legal framework within which AI technologies must operate.[8] Section 2 of the Constitution stipulates that any law or conduct inconsistent with the Constitution is deemed invalid. Organisations developing AI systems must ensure that these technologies align with constitutional values and fundamental rights.
II)Section 9- The Right to Equality
Section 9, titled “The Right to Equality” ensures that all individuals are treated fairly and equally, irrespective of their status, race, gender. [9]It also bans unfair discrimination against any individual. This right is particularly relevant to AI, as many automated systems depend historical data that mirrors existing social and economic inequalities. AI systems can inadvertently discriminate against specific groups; for example, an AI trained on biased historical hiring data might systematically disadvantage a black female applicant, thereby violating the constitutional right to equality. The Constitutional Court has affirmed that equality entails both eradicating unfair discrimination and advancing substantive equality[10] thereby raising critical legal questions about indirect discrimination and the accountability of organisations employing AI systems AI systems for discriminatory algorithmic outcomes.[11]
III)Section 10- The Right to Human Dignity
Section 10 guarantees that everyone individual possesses the right to human dignity. [12]The use of AI has raised concerns about this value because automated decision-making reduces individuals to statistical profiles rather than recognising their unique personal circumstances. Decisions concerning employment, education, and access to financial services may deprive individuals of meaningful human engagement and procedural fairness. The Constitutional Court has ruled that respect for human dignity is the foundation of every constitutional right and must remain central to the interpretation and application of legislation.[13]
IV)Section 14- The Right to Privacy
Section 14 of the Constitution safeguards every individual’s right to privacy, including protection from the unlawful collection and dissemination of personal information.[14] Many AI systems frequently rely on the collection, storage, analysis and sharing of data and personal information. The constitutional right to privacy mandates that personal information be collected and processed lawfully, reasonably, and used fairly. AI systems that gather excessive data and monitor individuals without proper consent violate constitutional privacy rights. AI applications such as facial recognition, biometrics, and identification systems frequently process highly sensitive personal data. This consequently heightens the risk of surveillance, identity theft and unauthorised data breaches. The Constitutional Court has acknowledged that privacy encompasses not only physical intrusion but also informational privacy, particularly in today’s increasingly digital society.[15]
V)Section 34- Access to courts and procedural fairness
Section 34 ensures that all individual’s have the right to have their legal disputes resolved through fair public hearings before independent courts. The intensive use of AI has sparked concerns about procedural fairness and access to justice. People impacted by automated government decisions have the right to understand how these decisions were made and to challenge them.
C) Protection of Personal Information Act 4 of 2013
The POPIA Act is currently the most significant legislation impacting AI in South Africa, as artificial intelligence relies on the collection, storage and processing of personal information. [16] The primary challenge is that modern AI systems require vast quantities of data to improve their accuracy. Consequently, organisations gather more data than is reasonably required, leading to significant risks of excessive surveillance and misuse of personal information.
POPIA also mandates that responsible parties implement suitable security measures to safeguard personal information from unauthorised access and cyberattacks. However, POPIA was not specifically drafted to regulate artificial intelligence. While it fails to tackle concerns like automated decision-making and algorithmic transparency, it does provide valuable safeguard for personal information.
D) The Cybercrimes Act 19 of 2020
This act criminalises various forms of unlawful conducts involving computer systems and electronic communication. This is highly relevant because AI technologies both enable and combat cybercrime. Criminals are increasingly leveraging AI to create phishing attacks, fabricate social media profiles, and automated hacking and identity theft schemes. Organisations employ AI-driven cybersecurity tools to detect malicious activities, pinpoint network vulnerabilities and respond to cyber threats.
E) The Legal Significance of Artificial Intelligence in South Africa
The legal implications of AI have transcended mere technological progress, directly impacting constitutional rights, liability and public trust. As AI systems increasingly influence decisions affecting individuals, the law must evolve to ensure transparency, fairness and accountability. While AI has driven economic growth and transformed numerous sectors of society, however, it has also introduced legal complexities that challenge established legal principles.
I)Artificial Intelligence and Employment Law
The workplace has emerged as one of the most rapidly expanding arenas for AI use. Employers utilise AI- driven software for recruiting staff, monitor workplace performance and even suggesting dismissals. This technology enhances efficiently by reducing administrative burdens, proving cost-effective and time-saving while enabling organisations to process large volumes of information quickly.
However, the use of AI in the workplace raises substantial legal issues under the Labour Relations Act 66 of 1995, the Employment Equity Act 55 of 1998 and the constitutional right to equality. Recruitment algorithms trained on historical employment data may unintentionally discriminate against candidates based on gender, race, disability.[17] Such outcomes constitute indirect discrimination even though it was unintentional. Employees facing automated disciplinary decisions may be denied procedural fairness even when no significant harm is present. South African Labour Law mandates that employers adhere to fair procedures before dismissing employees. It was argued that employers must continue to bear legal responsibility for employment decisions made with the assistance of AI. In Sidumo v Rustenburg Platinum Mines Ltd , the Constitutional Court affirmed that fairness is the foundation of South African Labour Law and that employment decisions must be evaluated against constitutional values.[18]
II)Artificial Intelligence and Healthcare
AI systems have also proven highly beneficial in enhancing service delivery within the healthcare sector. AI systems assists medical practitioners by analysing X-rays, predicting patient outcomes, managing administrative records, and supporting treatment plans. These innovations have proven valuable especially in rural areas where medical services are very scarce. However, AI- assisted healthcare services raise challenging questions regarding professional negligence and liability. For example, if an AI diagnostic system was to produce an incorrect diagnosis that results in patient harm, determining legal responsibility becomes very hard. Liability might go to the medical practitioners, hospital, software developer of the AI system. South African medical negligence law has focused on the conduct of healthcare professionals. AI complicates this approach because medical decisions involve collaboration between human practitioners and intelligent software. Section 27 of the Constitution guarantees everyone the right to have access to basic healthcare services. [19]The Constitutional Court held that the state bears a constitutional obligation to take reasonable measures to realise this right progressively.[20]
III)Artificial Intelligence and Criminal Justice
Artificial intelligence has begun to influence the criminal justice system. AI technologies has assisted law enforcement agencies by identifying criminal patterns, analysis of forensic evidence, detection of cybercrime and cyberattacks and improving investigative efficiency. [21]However, predictive policing algorithms also present serious constitutional concerns, where historical crime data reflects existing inequalities. AI systems may reinforce discriminatory policing practices by disproportionately targeting certain communities. Such outcome risks undermine the constitutional principles of equality, dignity and fairness.
Similarly, AI-generated evidence can raise questions regarding reliability, transparency and credibility in criminal proceedings. Courts will then have to remain cautious when replying upon algorithmic evidence that cannot easily be independently verified.
F) Comparative Perspectives
I)The European Union: The AI Act
The European Union has adopted the first comprehensive legal framework regulating artificial intelligence through the Artificial Intelligence Act. This act adopts a risk-based approach, categorising AI systems according to the level of risk they pose to individuals and society. In this act, the AI systems were classified into four categories: (a)unacceptable-risk AI, (b)High-risk AI including systems used in healthcare, employment, education, law enforcement agencies, (c)Limited-risk AI which must comply with transparency obligations and (d) Minimal-risk AI.[22]
The act requires developers and deployers of high-risk AI systems to conduct risk assessments, maintain detailed technical documentation and demonstrate compliance before placing AI products on the market. This framework reflects the principle that the greater the potential harm posed by AI, the greater the legal obligations imposed upon those responsible for developing and executing it. South Africa could benefit significantly from adopting a similar risk-based model, in sectors such as healthcare, public administration, financial services where AI decisions directly affect constitutional rights.
II)The United Kingdom
The United Kingdom had a more flexible and sector-specific approach to AI regulations. Its regulatory model was guided by five principles: (a) safety and security, (b) fairness, (c) transparency and explainability, (d)Redress, (e) accountability and governance.[23] This approach allowed regulators to adapt to AI governance according to the particular risks while avoiding unnecessary barriers to innovation. South Africa currently resembles the UK approach because AI-related issues are addressed indirectly through existing legislation.
G) Critical Evaluation: Is South Africa’s Legal Framework Adequate?
Although South Africa possesses a strong constitutional foundation and several statutes that indirectly regulate artificial intelligence, the current legal framework remains fragmented and reactive. Existing legislation was enacted before the rapid emerging of AI technologies and therefore addresses only isolated aspects of AI governance rather than providing a comprehensive regulatory framework.
One of the most significant shortcomings is the absence of legislation specifically implemented to govern automated decision-making. While POPIA regulates the processing of personal information, it provides limited guidance regarding algorithmic transparency or accountability where AI systems make decisions that significantly affect individuals. Another weakness lies in determining legal liability when AI systems cause harm. Legal principles assume that human beings make decisions and can be held accountable for unlawful conduct, while with AI systems it is difficult to detect who will carry the legal liability. Without dedicated legislation, courts may experience difficulty in identifying the appropriate party to bear legal responsibility.
South Africa currently lacks a specialised regulatory authority responsible for overseeing AI governance. Different aspects of AI fall under various regulators, including the Information Regulator, the sector-specific bodies. This fragmented oversight creates uncertainty for businesses, developers and consumers. Despite these weaknesses, South Africa is well positioned to develop an effective AI regulatory framework. The constitution provides a strong human rights foundation and the existing legislation demonstrates a commitment to protecting privacy, equality, dignity and access to justice.
H) Recommendations
The increasing advancement of AI has presented South Africa with a huge opportunity to become a leading country in responsible AI governance on the African continent. The first recommendation is to enact a comprehensive AI legislation.
South Africa should enact a comprehensive AI legislation that will clearly define artificial intelligence, establish the different categories of AI systems based on risk and outlines the legal responsibility for developers and users if these systems cause harm or discrimination. Such legislation will provide certainty regarding liability, regulatory compliance and enforcement while remaining flexible to accommodate future innovations. A risk-based regulatory model similar to the European Union would ensure that artificial intelligence systems that pose threats to constitutional rights are subjected to stricter legal obligations.
I)Promote Transparency and Accountability
Artificial intelligence systems that affect individuals should be capable of providing understandable explanations for their decisions. Individuals should have the right to know when an automated system has been used to make decisions relating to employment, education, healthcare, and public administration. Greater transparency will enhance public trust in AI technologies while improving accountability and enabling individuals to challenge unlawful or unfair decisions.
II)Establish a National Artificial Intelligence Regulatory Authority
South Africa should create a specialised regulatory body that will be responsible for overseeing AI governance. Such authority will be able to monitor compliance with the AI legislation, issue regulatory guidance, investigate complaints and advise Parliament on emerging technological developments. Having a dedicated regulator would improve consistency and provide businesses, government institutions and consumers great legal certainty.
III)Invest in Artificial Intelligence Education and Capacity Building
Improved legal and technical expertise support will be a good course and will allow the country to develop a responsive legislation that will be capable of adapting to future technological innovations. A effective AI regulation will require legal practitioners, policymakers and regulators to understand the capabilities and limitations of AI technologies. Universities and professional bodies should therefore expand education and training programmes that will fully focus on technological law, AI ethics, cybersecurity, cyberattacks and digital governance.
Conclusion
South Africa has the potential to leverage AI to significantly enhance healthcare, improve financial services, modernise public administration, enhance education and drive economic growth. This technological system has become one of the key innovations of our time, fundamentally transforming the way government, businesses, and individuals engage, be it through virtual meetings or general business proposals. However, these advantages come with considerable legal and ethical issues, including concerns related to privacy, equality, accountability, transparency, cybersecurity and liability.
This article illustrated that while South Africa’s Constitution and its existing legislation, such as the Protection of Personal Information Act, the Cybercrimes Act, and the Electronic Communications and Transactions Act, lay a vital legal foundation, they do not adequately address the unique challenges presented by artificial intelligence.
The comparative analysis of jurisdictions such as the European Union and the United Kingdom illustrated that governments are actively developing legal frameworks that will be capable of balancing innovation with human rights protection. The future of AI governance in South Africa will depend upon the country’s ability to encourage innovations while ensuring that technological advancements remain consistent with the principles of dignity, equality, accountability, transparency and the rule of law. By adopting comprehensive legislation and strengthening institutional insight, the country can be able to position itself as a responsible leader in AI governance, ensuring that technological progress serves the interests of all members of society.
Reference(S):
Legislation
-Constitution of the Republic of South Africa, 1996
-Protection of Personal Information Act 4 of 2013
-Cybercrimes Act 19 of 2020
-Electronic Communications and Transactions Act 25 of 2002
-Labour Relations Act 66 of 1995
-Employment Equity Act 55 0f 1998
– Promotion of Equality and Prevention of Unfair Discrimination Act 4 of 2000
-Artificial Intelligence Act (European Union) 2024
Case Law Analysis
National Media Ltd v Jooste 1996 (3) SA 262 (A)
Khumalo v Holomisa 2002 (5) SA 401 (CC)
NM v Smith 2007 (5) SA 250 (CC)
Minister of Health v Treatment Action Campaign 2002 (5) SA 721 (CC)
Investigating Directorate: Serious Economic Offences v Hyundai Motor Distributors (Pty) Ltd 2001 (1) SA 545 (CC)
Harksen v Lane NO 1998 (1) SA 300 (CC0
S v Makwanyane 1995 (3) SA 391 (CC)
Sidumo and Another v Rustenburg Platinum Mines Ltd and Others 2008 (2) SA 24 (CC)
Journal Articles
Karen Yeung, ‘A study of Algorithmic Regulation’ (2018) 12 Regulation and Governance 505
International AI authorities
European Union AI Act (Regulation (EU) 2024/1689
UNESCO Recommendations on the Ethics of AI (2021)
[1] Department of Communications and Digital Technologies, South Africa National Artificial Intelligence Policy Framework (2024)
[2] UNESCO, Recommendations on the Ethics of Artificial Intelligence (2021)
[3] Protection of Personal Information Act 4 of 2013; Electronic Communications and Transactions Act 25 of 2002; Cybercrimes Act 19 of 2020
[4] Constitution of the Republic of South Africa, 1996
Section 9- Equality
Section 10-Human dignity
Section 14-Right to privacy, section 34-Access to courts and Procedural fairness
[5] S v Makwanyane 1995 (3) SA 391 (CC) para 144
[6] S v Makwanyane 1995 (3) SA 391 (CC) para 144; Harksen v Lane NO 1998 (1) SA 300 (CC)
[7] Constitution of the Republic of South Africa, 1996; Protection of Personal Information Act 4 of 2013; Electronic Communications and Transactions Act 25 of 2002; Cybercrimes Act 19 of 2020
[8] Constitution of the Republic of South Africa, 1996 s9
[9] Constitution of the Republic of South Africa, 1996 s9
[10] Harksen v Lane NO 1998 (1) SA 300 (CC) para 46-53
[11] Karen Yeung, ‘A study of Algorithmic Regulation’ (2018) 12 Regulation & Governance 505
[12] Constitution of the Republic of South Africa 1996, s10
[13] S v Makwanyane 1995 (3) SA 391 (CC) para 144-146
[14] Constitution of the Republic of South Africa, 1996 s14
[15] Investigating Directorate: Serious Economic Offences v Hyundai Motor Distributors (Pty) Ltd 2001 (1) SA 545 (CC) para 15-18; NM v Smith 2007 (5) SA 250 (CC) para 131-136
[16] Protection of Personal Information Act 4 of 2013
[17] Employment Equity Act 55 of 1998; Constitution of the Republic of South Africa, 1996 s9
[18] Sidumo v Rustenburg Platinum Mines Ltd and Others 2008 (2) SA 24 (CC) para 74-79
[19] Constitution of the Republic of South Africa, 1996 s27
[20] Minister of Health v Treatment Action Campaign 2002 (5) SA 721 (CC) para 94-95
[21] Cybercrimes Act 19 of 2020; OECD, Recommendation of the Council on Artificial Intelligence (2019)
[22] Regulation (EU) 2024/1689, European Union Act
[23] UK Government, A Pro-Innovation Approach to AI Regulation (Department for Science, Innovation and Technology, 2023
