Artificial intelligence systems trained largely on foreign datasets risk perpetuating racial bias and undermining South Africa’s transformative constitutional vision if they are not built with local historical context, Justice and Constitutional Development Minister Mmamoloko Kubayi has warned.
Kubayi delivered the warning on Thursday at the University of Johannesburg’s “AI and the Law” conference, arguing that while AI-powered tools could sharply reduce legal costs and delays, the same technology could just as easily widen inequality if deployed without transparency, accountability and data that reflects the country’s apartheid past.
“African data which carry our archives, African languages and cultural nuance, do not find expression in the AI machines that are currently used in the various fields, including law,” Kubayi said. “Most AI is trained on datasets from North America and Asia. This exclusion perpetuates data poverty and algorithmic bias, leading to real-world consequences such as poor language translation machines, especially for African languages.”
The bigger risk for South Africa, she said, was the use of AI systems “totally devoid of our historical context as a country.” The minister linked that gap directly to the constitutional project, arguing that South Africa’s Constitution is transformative and was drafted to heal past divisions and lay a foundation for social justice.
“As a country we have seen and experienced the horrific consequences of a nation organised through racial segregation,” Kubayi said. “We need to ask ourselves: what is it that we need to do [so] that technological progress is not concomitant with the growth of digitised racial segregation?”
She pressed the audience further: “The data that carries our history and our social context has been excluded in the development of the AI machines that are currently in the market. How then should we expect that the machines which are, in all likelihood, biased against us will be able to help us build a just society? Are we not, in our quest to increase access to justice through AI, at the risk of increasing access to injustice?”
Kubayi acknowledged, however, that AI could meaningfully broaden access to justice if the underlying data problem were addressed. She cited prohibitive legal costs, geographic remoteness from courts and legal services, lengthy delays in case resolution, and a shortage of legal practitioners willing or able to serve lower-income communities as long-standing barriers to universal access to justice.
“By automating repetitive processes, advancing legal research, drafting, increasing dispute resolution, and encouraging predictive legal analytics, the introduction of AI-powered solutions has drastically transformed the delivery of legal services,” she said. “The effect of AI-powered tools does not only reduce the time it takes to perform these tasks, but they also do these tasks at a fraction of the traditional cost.”
The minister called on universities to help work through those tensions, asking the conference to interrogate how AI can be harnessed to advance the Sustainable Development Goals through law, governance and practice, and said she looked forward to the outcomes of the deliberations.
