South Africa’s Financial Sector Conduct Authority will wait for the Financial Stability Board to publish global principles on AI adoption before developing its own rules, commissioner Unathi Kamlana said.
The Basel-based FSB is finalising guidance to help financial firms adopt AI responsibly, listed as a G20 deliverable under the 2026 US presidency and due for publication next month.
“It’s good for us to be waiting for that because we want to be aligned,” Kamlana said in an interview on Friday.
The FSB principles will inform a joint discussion paper being prepared by the FSCA, the Prudential Authority and the South African Reserve Bank, which will include specific considerations on agentic AI and lay groundwork for a future regulatory framework.
Principles, not rules
Kamlana was clear that the regulators do not intend to write prescriptive rules. “We are going to stay the course on AI and the risk it presents to the financial sector, but we have been clear and consistent on the approach, which is principles rather than rules, because it is a fast evolving area,” he said.
That places South Africa closer to the UK and Singapore than to the European Union, whose AI Act classifies systems by risk tier and attaches binding obligations to each.
The FSCA and Prudential Authority have already published a market scan of AI adoption in the financial sector, which Kamlana described as an examination of dominant use cases rather than a governance document. That study found AI adoption at 52% among banks and 50% among fintechs, but just 8% among insurers — the most cautious segment of the market. It also noted that no uniform AI governance framework exists in South Africa, and that international standards such as the OECD principles and the EU AI Act carry no binding force locally.
A subsequent paper will set out how regulators intend to incorporate governance principles into financial institutions’ use of AI.
Why agentic AI matters here
The decision to address agentic AI specifically is the most forward-looking element.
Agentic systems act rather than advise — executing transactions, moving funds and making decisions with limited human review. That shifts the regulatory question from whether a model’s output is accurate to who is accountable for what an autonomous system does.
South African banks are already deploying such systems. Absa is testing three autonomous agents through its partnership with Salesforce’s Agentforce platform. The BioCatch Future of Digital Trust survey found 86% of South African banking leaders identify AI agents as the industry’s most exploitable vulnerability over the coming year. Boston Consulting Group estimated this month that agentic AI could cut the cost of running scams by 90% or more.
Three processes, all waiting
The FSCA’s position is defensible on its own terms. South Africa’s largest banks operate internationally, and rules misaligned with global standards would impose compliance costs without clear benefit.
It does, however, join a pattern. South Africa’s national AI policy was withdrawn in April after its reference list was found to contain fabricated citations; a redraft goes to Cabinet in November, with public consultation targeted for January 2027. President Cyril Ramaphosa used last week’s BRICS summit to propose an international mechanism for independent scientific evaluation of AI.
Elsewhere on the continent, regulators are moving on different timetables. Egypt’s Financial Regulatory Authority has run an AI regulatory sandbox since November 2024, admitting seven projects in its first year. Nigeria’s AI Control and Regulation Bill names credit scoring as a high-impact decision requiring disclosure and a route to challenge. Mozambique’s education ministry expects AI regulation at Cabinet before year end.
South Africa is choosing alignment over speed. Whether that proves the more durable approach will depend on how quickly the FSB report arrives — and how long the discussion papers that follow it take.
