AI already decides who gets a loan in Kenya. Now decide who governs it

AI models already determine loan approvals in Kenya, without human credit officers reviewing applications. The European Union has set rules for AI in credit scoring, classifying it as high-risk and imposing strict requirements. However, Kenya should not follow a one-size-fits-all approach. The country can learn from its own experience of scoring without governance, resulting in many Kenyans being blacklisted on credit bureaus.

To address this issue, the Digital Financial Services Association of Kenya suggests a model governance framework. Model governance beats model bans; regulations should focus on how AI systems are used, not the technology itself. Lenders should demand specific information from vendors about the data used, model error rates, bias testing, and audit logs.

A named human owner for every model in production is required, along with plain-language explanations for automated declines. The Central Bank should lead the regulatory effort, with lenders keeping a register of models and providing borrowers with clear reasons for automated declines. This framework can be adapted to Kenya's market, avoiding unnecessary bureaucracy and enabling the industry to continue its growth. Clear rules are more cost-effective than facing scandals.