South Africa Holds Off on AI Rules as Financial Regulators Wait for Global Principles

South Africa’s Financial Sector Conduct Authority is holding back from writing dedicated artificial-intelligence rules for the financial sector until the Financial Stability Board publishes a set of global principles. The regulator’s approach is to align with international guidance while developing a framework that reflects the country’s own market and risks.

FSCA commissioner Unathi Kamlana said the FSB guidance is expected to inform a joint discussion paper being prepared with the Prudential Authority and the South African Reserve Bank. That paper is expected to address the use of AI in financial institutions, including the governance questions raised by systems that can act with limited human review.

The decision places South Africa in a principles-based regulatory camp. Kamlana said the authorities do not intend to prescribe a detailed rulebook for every use case because the technology is changing quickly. The regulators are instead expected to define expectations around risk, accountability and responsible adoption, then apply those principles as the market develops.

That choice is consequential because adoption is already uneven across the financial system. A previous market scan found AI being used by 52% of banks and 50% of fintechs, compared with 8% of insurers. The same work found that South Africa does not yet have a uniform AI-governance framework with binding force, leaving firms to interpret international standards and sectoral expectations for themselves.

The most forward-looking issue is agentic AI. Unlike systems that simply generate advice or content, agentic tools can execute transactions, move money or make decisions with limited human intervention. That changes the regulatory question from whether a model’s output is accurate to who is accountable for the system’s actions, how those actions can be audited and what recourse customers have when something goes wrong.

South Africa is therefore waiting on three linked processes: the FSB’s global principles, the domestic discussion paper and the country’s wider national AI policy work. Waiting can reduce the risk of incompatible rules, but it also leaves banks, insurers and fintechs operating in a period of uncertainty.

The durability of the approach will depend on speed after the global guidance arrives. Alignment is useful only if it is followed by clear local expectations, credible supervision and practical guidance that institutions of different sizes can implement.