South Africa: Draft National AI Policy Withdrawn Amid Governance Concerns

• South Africa ranks eleventh in Africa for AI governance capacity, despite leading the continent in AI infrastructure and adoption.
• The Africa AI Governance Index, published by the Lawyers Hub, scores South Africa at 2.10 points, significantly behind leaders like Rwanda (3.25), Nigeria (2.81), and Benin (2.58).
• The Draft South Africa National AI Policy, published on April 10, 2026, was withdrawn on April 26, 2026, by Minister Solly Malatsi after News24 revealed "AI-hallucinated references" in its academic citations.
• The withdrawal, gazetted on June 12, led to the suspension of two officials and left South Africa with no dedicated national AI policy.
• The Department of Communications and Digital Technologies plans to submit revised AI policy documents to Cabinet by November 2026, targeting public comment by January 2027.

South Africa currently operates without any dedicated national artificial intelligence policy, creating a significant regulatory void.

South Africa has long been recognized as a technological leader on the African continent, boasting significant digital infrastructure including cloud regions hosted by global giants like Google, Microsoft, and Amazon Web Services. Its major banks are also among the most advanced adopters of artificial intelligence (AI) across Africa. However, a recent continental assessment reveals a stark contrast between this technological prowess and the nation's capacity for AI governance.

A new continental index, published in July, places South Africa eleventh in Africa for its ability to govern AI. This ranking puts it behind ten other nations, including Rwanda, Nigeria, Benin, Kenya, Morocco, Egypt, Ethiopia, Ghana, Mauritius, and Tunisia, and on par with Zimbabwe. The Africa AI Governance Index, developed by the Nairobi-based Lawyers Hub, comprehensively evaluates all 54 African states across eight weighted pillars, ranging from strategy and regulation to infrastructure, human capital, ethics, implementation, and impact. The index's methodology is robust, relying on 80 indicators per country, compiled by 97 policy fellows, and supported by over 4,000 evidence-linked observations, with all country workbooks openly published for scrutiny. This assessment, grounded in the African Union's Continental AI Strategy, was launched in Geneva at the inaugural UN Global Dialogue on AI Governance, in collaboration with South Africa's own Human Sciences Research Council.

The disparity is particularly concerning given South Africa's high exposure to AI. While Rwanda leads the index with 3.25 points out of four, followed by Nigeria at 2.81 and Benin at 2.58 – the only three states classified as having established AI governance – South Africa scores a modest 2.10. Africa as a whole accounts for less than 1% of global data centre capacity despite housing 18% of the world's population, with existing capacity heavily concentrated in South Africa, Egypt, Kenya, Nigeria, and Morocco. This concentration means South Africa carries more inherent risk from unmanaged AI than many of its continental peers. For instance, Google announced in July at its Africa Cloud Summit in Johannesburg that it had surpassed its $1 billion African investment commitment, while South African banks and retailers extensively deploy AI for critical functions such as credit assessments, pricing strategies, and customer decision-making on a national scale.

The immediate cause of South Africa's precarious position in AI governance is well-documented and recent. On April 10, 2026, the Department of Communications and Digital Technologies (DCDT) released the Draft South Africa National AI Policy for public comment. This initiative, intended to establish a framework for responsible AI use, quickly ran into significant issues.

Within two weeks of its publication, investigative reporting by News24 revealed that several academic sources cited within the draft document did not actually exist. This critical flaw, involving as many as six "AI-hallucinated references," led to the swift intervention of Minister Solly Malatsi, who formally withdrew the policy on April 26. The withdrawal was subsequently gazetted on June 12, and two officials were placed on precautionary suspension in connection with the incident.

This episode highlighted an uncomfortable irony: a policy designed to govern the responsible application of AI was itself undermined by the unverified use of AI in its drafting process. The failure was not technological in nature but rather a lapse in due diligence, indicating a lack of proper verification during the policy's creation.

Following the South Africa Draft National AI Policy withdrawal, the nation currently finds itself in a significant SA AI regulation vacuum, operating without any dedicated national artificial intelligence policy. This absence creates a challenging environment for businesses and legal professionals navigating the rapidly evolving AI landscape.

The Department of Communications and Digital Technologies (DCDT) has outlined a recovery path to address this policy void. Revised documents are slated for submission to Cabinet by November 2026, with a target for public comment in January 2027. Minister Malatsi has acknowledged the imperative to minimize a prolonged policy vacuum, signaling an understanding of the urgency, though this also represents a concession regarding the initial misstep.

This situation in South Africa mirrors a broader trend observed across the continent. The Africa AI Governance Index indicates that while approximately 15 African states have adopted national AI frameworks and another 24 are in the drafting stages, few have allocated ring-fenced budgets, established independent oversight bodies, or implemented effective mechanisms to monitor the fulfillment of their AI commitments. Across every African sub-region, strategy consistently scores highest, while implementation lags significantly, underscoring a continent-wide challenge in translating policy intent into practical governance.

This article highlights a critical regulatory vacuum in South Africa's AI governance, following the withdrawal of the flawed draft national policy. Lawyers and compliance officers must advise clients on operating in an environment with significant AI exposure but without dedicated legislation, and closely track the Department of Communications and Digital Technologies' revised policy timeline for future compliance obligations.

Source: Reporting based on analysis from AIforBusiness.net.