Abstract
This dissertation critically evaluates whether artificial intelligence (AI) can or should be
recognised as a legal subject within the South African legal system. As AI systems increasingly
influence decisions across both public and private domains, their growing autonomy raises
questions regarding accountability, liability, administrative justice and the conceptual
boundaries of legal subjectivity. Using a doctrinal legal methodology supported by a focused
comparative analysis, the study interrogates whether AI possesses, or could be said to possess,
the theoretical, philosophical and juridical attributes required for legal subjectivity in South
Africa. Foundational concepts of natural and juristic personhood are examined, demonstrating
that legal subjectivity is grounded in intention, consciousness, moral agency and
accountability, qualities that contemporary AI systems do not possess. The analysis then
evaluates how South Africa’s constitutional, statutory and administrative frameworks respond
to automated decision-making, drawing on key legislation including the Protection of Personal
Information Act No. 4 of 2013 particularly section 71, the Electronic Communications and
Transactions Act No. 25 of 2002 the Promotion of Administrative Justice Act No.3 of 2000
and the Constitution of the Republic of South Africa 1996.
The study further incorporates contemporary academic scholarship and considers Kenya’s Data
Protection Act, as well as judicial developments such as Okiya Omtatah Okoiti vs
Communications Authority of Kenya & 8 others [2018] eKLR to offer relevant comparative
insights. These comparative findings reveal a shared commitment to human-centred
accountability and demonstrate that robust regulation of AI does not require extending legal
personhood to autonomous systems. The South African analysis also identifies several
regulatory gaps relating to algorithmic opacity, evidentiary difficulties, enforcement capacity
and liability diffusion, underscoring the need for a more coherent governance framework.
The dissertation concludes that AI cannot meet the ontological or functional requirements for
legal subjectivity under South African law. Rather than extending personhood to AI, South
Africa should strengthen human-centred liability frameworks, enhance transparency and
oversight mechanisms and consider developing a dedicated AI governance framework.