According to a doctoral dissertation from the University of Cape Town, South Africa should regulate artificial intelligence through mandatory rules enshrined in the constitution, rather than relying on voluntary ethical principles and self-regulation.
Hybrid Model
This argument is put forward by lawyer and researcher Nokutula Olorunju amid the Ministry of Communications and Digital Technologies revising national AI policy with a deliberately soft approach. Minister Sulli Malatsi has tied the future of AI in South Africa to maintaining neutrality between Washington and Beijing, and the new draft establishes general guidelines while leaving strategy development to separate sectors.
Olorunju does not comment on the ministry's work, but her findings clearly contradict this course. In her dissertation, titled 'Artificial Intelligence, Bias and the Rule of Law: A Constitutional Approach to the Regulation of AI in South Africa,' she argues that existing legislation, non-binding ethical principles, and self-regulation are insufficient. In her view, principles lack legal force, and self-regulation allows organizations to choose the easiest commitments to include in their missions.
She notes that fragmentation, weak enforcement, and legal barriers contribute to the spread of misuse, creating overlapping jurisdictions and uncertainty about which body is responsible. 'Essentially, if there is a loophole, it will probably be exploited,' she stated.
Nokutula Olorunju, a research fellow at Research ICT Africa, will defend her dissertation at UCT on Tuesday, September 8. Her comments appeared in the UCT News graduate profile published on September 2, as well as in a university statement.
Hybrid Model
Her proposal is a constitutionally based hybrid model: conflicting mandates must be clarified, an interdisciplinary AI oversight body must be created, and duplication of governance structures must be avoided. Developers and operators must demonstrate compliance with human rights standards and assume responsibility throughout the system's lifecycle. In her opinion, compliance should be more significant than annual formal confirmation.
The university indicates that for ordinary users, this means introducing clearer rules governing AI tools, AI system assessments, accountability measures, and procedures for filing complaints about AI misuse. Olorunju also insists on the availability of informational materials to raise public awareness in all 12 official languages.
Existing institutions and industry regulators can take on part of the burden provided they are properly equipped and their powers are clearly defined. Nevertheless, this does not make her a simple critic of South Africa's delays. UCT presents her arguments as a call to move away from a 'long-term waiting approach.' She is more reserved in her profile.
'This approach has two sides,' she told UCT News. 'On one hand, South Africa has benefited from observing how other countries approach AI regulation; on the other hand, waiting too long could lead to adopting AI laws on a 'copy-paste' basis that are irrelevant to the South African context.'
Malatsi withdrew the AI policy draft in April after discovering references to non-existent academic sources, subsequently appointing an independent commission chaired by Wits AI researcher Benjamin Rosman to develop a replacement with a softer approach, which is due to be presented to the cabinet later this fiscal year. This delay has already resulted in a loss of visibility, as South Africa is absent from the World Bank's count of over 80 countries with published AI strategies.
Call for International Experience
Olorunju argues that South Africa should rely on best international practices, rather than fully adopting regulatory models from the US or EU, and since there is no perfect way to regulate AI, countries should start with what they can realistically implement.