Governing AI in emerging financial markets

The most interesting AI governance experiment in the world right now is not happening in developed economies, rather some crucial experiments are happening in developing economies including India, Indonesia and Kenya. Emerging economies are facing a unique challenge. They must regulate rapidly expanding AI-driven financial systems that serve populations with limited legal recourse, often without the institutional capacity assumed by Western regulatory frameworks. This has led them to develop governance approaches that, in several respects, are more sophisticated and better adapted to their context than the models they are typically compared against.

Why copy-pasting doesn’t work

The EU AI Act is a significant piece of legislation designed for the institutional and social realities of developed economies. It reflects a context with strong regulatory institutions, well-established legal systems, mature supervisory capacity, and relatively integrated financial markets. It also focuses upon social priorities, including welfare and environmental protection, that have evolved through Europe’s own historical and institutional development. Expecting emerging economies to adopt the same risk-classification framework without adaptation overlooks these differences. It is similar to requiring a banking system that is still developing basic credit infrastructure to fully implement advanced prudential standards such as Basel III. The framework itself may be robust, but it is not automatically suited to a different stage of institutional development.

Contrary, the US approach presents a different challenge. Instead of a single, comprehensive AI law, it relies on sector-specific guidance from financial regulators and consumer protection agencies, with enforcement often taking place after harm has occurred. This approach works in a system with strong legal institutions, active litigation, and effective mechanisms for consumer redressal.

In many emerging economies, however, these conditions do not exist to the same extent. Consumers often have limited access to legal remedies, while information asymmetries between financial institutions and individuals remain significant. When AI systems deny access to credit, insurance, or other essential financial services, waiting for risk to occur before taking action can leave vulnerable populations without meaningful protection. As a result, a purely retrospective, enforcement-led approach is often insufficient.

Faced with these realities, regulators in many emerging economies have, largely out of necessity, begun to develop a third approach; one that places greater emphasis on preventive governance, institutional oversight, and context-specific safeguards rather than relying solely on either comprehensive legislation or ex-post enforcement.

A practical customised framework

The experience of emerging economies shows that effective AI governance does not require choosing between the European and American models. Instead, it requires building a framework that reflects local institutional realities while meeting relevant global standards. This framework has two complementary pillars: 

1. Strong governance within financial institutions 
2. Effective oversight by regulators

Within financial institutions, AI can no longer be treated as a technology project or a compliance issue. It has become a core business and risk function. AI systems increasingly decide who receives credit, how risks are priced, which transactions are flagged as suspicious, and how investment decisions are made. As a result, responsibility for AI governance must sit at the highest levels of management.

Leading institutions are moving in this direction. They are assigning clear executive accountability for AI, just as financial institutions established Chief Risk Officers after earlier financial crises. They are also ensuring that important decisions made by AI remain subject to meaningful human oversight. This means having people who can review, challenge, or stop an algorithm when necessary; not simply documenting procedures that are never tested in practice.

Another important lesson is that while technology can be outsourced, accountability cannot. Financial institutions may rely on external vendors for AI systems, but they remain responsible for the outcomes. Contracts with technology providers should therefore include clear rights to audit AI systems, assess their performance, and address governance failures.

Regulators, meanwhile, are increasingly focusing on three areas.

• First, AI decisions must be explainable.
• Second, fairness must be assessed before deployment, not after problems emerge. 
• Finally, customers must have meaningful avenues for redress. 

Together, these measures represent a practical approach to AI governance. They recognise that emerging economies cannot simply copy regulatory models developed elsewhere. Instead, they are building governance systems that reflect their own institutional capacities, financial inclusion priorities, and development needs while ensuring that AI remains transparent, accountable, and fair.

Learning outcome for leadership and board

1. AI governance cannot remain delegated. The materiality of algorithmic decision-making to balance-sheet risk and consumer outcomes has crossed the threshold at which board-level attention is a regulatory expectation, not a best practice recommendation.
2. Third-party model risk is probably underpriced. Institutions that have built AI capabilities on vendor-supplied models without independent validation, without audit rights, and without tested suspension protocols are carrying exposures that may not be accurately reflected in their risk frameworks. This is not a future concern. It is a current position.
3. And regulatory expectations will intensify, not moderate. Institutions that treat current compliance requirements as the ceiling of their obligation will find themselves continuously reactive. Those that build genuine governance capability as operational competencies will be better positioned across the regulatory cycle and, increasingly, in competitive terms.

The global debate about whether to regulate AI in financial services is largely settling down, but the complex step on HOW? Is still unanswered due to lack of research and credible models available in the market.

END OF ARTICLE