The use of algorithmic credit scoring presents opportunities and challenges for lenders, regulators, and consumers. This paper provides an analysis of the perils of the use of AI in lending, such as the problem of discrimination in lending markets that use algorithmic credit scoring, the limited control financial consumers have over the outcomes of AI models due to the current scope of data protection law and financial consumer protection law, the financial exclusion caused by the lack of data from traditionally excluded groups, the regulatory arbitrage in lending markets, and the little oversight of the use of alternative data for algorithmic credit scoring. I provide a comparative overview of the current approaches to algorithmic credit scoring in different jurisdictions such as Kenya, the European Union, the United Kingdom, Hong Kong, Singapore, the United States, Australia, and Brazil to argue that these models do not solve the problems illustrated. To address the problems of algorithmic credit scoring and effectively protect consumers as end users of these models, and therefore, promote access to finance, this paper proposes a set of tools and solutions for financial regulators. First, a testing supervisory process for algorithmic credit scoring models will effectively promote fair lending. Second, to create a right to know the outcomes of the algorithm, including opinion data and inferences, to promote digital self-determination. This solution empowers consumers affected by algorithmic credit scoring so they can verify and challenge the decision made by the AI model. Third, to level the playing field between financial institutions and other lenders that use algorithmic credit scoring. Fourth, to use the sandbox as a test environment for lenders to create data of traditionally excluded groups in a controlled environment. And finally, to foster data sharing and data portability initiatives for credit scoring through open finance schemes in an environment controlled by the financial regulatory authority. Better algorithms, unbiased data, AI regulation, fair lending regulation and AI governance guidelines do not solve the perils of the use of AI for creditworthiness assessment. In contrast, these proposals aim to solve the problems of algorithmic credit scoring in any jurisdiction
