In a speech on regulation and supervision in the age of artificial intelligence, the South African Reserve Bank, through Deputy Governor Fundi Tshazibana in her capacity as Chief Executive Officer of the Prudential Authority, set out a three-part supervisory agenda for the financial sector focused on improving information, building skills and calibrating regulation. The clearest near-term policy step is a joint Prudential Authority and Financial Sector Conduct Authority discussion paper on the regulatory approach to AI, likely to be published early in the second half of 2026, with regulatory arrangements expected probably by early 2027 after stakeholder engagement. The speech frames AI as a growing prudential and conduct issue rather than announcing immediate rules. It highlights five vulnerabilities drawn from the Financial Stability Board monitoring framework: third-party dependencies, cyber risks, model risks, market correlations and misalignment. Supervisory priorities include developing a taxonomy that distinguishes between types of AI systems and risk levels of use cases, embedding AI into firms' risk management frameworks, keeping records on models and performance, using explainability techniques for high-impact decisions and requiring disclosure on how AI shapes those decisions. A recent Prudential Authority and Financial Sector Conduct Authority survey found AI is not yet deeply embedded across South African financial institutions, but identified shortages of AI skills and weaknesses in fairness testing, third-party model risk management, board-level oversight, executive accountability and operationalised ethical principles. The planned discussion paper will be the basis for further stakeholder discussions. In parallel, the Intergovernmental Fintech Working Group is developing an AI workstream, with sandbox testing expected to support more complex use cases such as algorithmic credit scoring, index insurance and tokenised assets.
South African Reserve Bank 2026-05-08
South African Reserve Bank outlines AI supervision agenda with joint discussion paper expected in early second half of 2026
The South African Reserve Bank’s Prudential Authority outlined a three-part AI supervisory agenda for the financial sector focused on improving information, building skills and calibrating regulation. A joint Prudential Authority–Financial Sector Conduct Authority discussion paper on AI regulation is expected in early H2 2026, with regulatory arrangements likely by early 2027. The speech highlighted AI as a growing prudential and conduct issue, citing vulnerabilities such as third-party dependencies, cyber and model risks, and prioritising AI taxonomies, governance, explainability and disclosure. It noted gaps in skills, fairness testing and board oversight. In parallel, the Intergovernmental Fintech Working Group is developing an AI workstream with sandbox testing for complex use cases including algorithmic credit scoring, index insurance and tokenised assets.