91

This article examines the role of artificial intelligence (AI) in credit markets. We focus on two types of models used to estimate the probability that borrowers will default : (1) origination models, applied when evaluating loan applications, and (2) internal models, developed in order to calculate banks’ regulatory capital requirements. These applications are critical: origination determines access to credit for households and SMEs, while internal models shape banks’ capitalization and, by extension, the stability of the financial system. While AI is now widely drawn on in origination, it remains as yet little used in regulatory models. The low level of adoption may be explained by the limited use of alternative data, modest economic incentives in terms of capital savings, and challenges related to model interpretability. JEL classification: C10, C38, C55, G21, G29.