The Safe-Tail Paradox: Stress Testing AI Exposure of Banks Borrowers

Christophe Hurlin () and Christophe Pérignon
Additional contact information
Christophe Hurlin: LEO - Laboratoire d'Économie d'Orleans [UMR7322] - UO - Université d'Orléans - UT - Université de Tours - NEOLAiA - NEOLAiA European University = Université Européenne NEOLAÏA - CNRS - Centre National de la Recherche Scientifique
Christophe Pérignon: HEC Paris - Ecole des Hautes Etudes Commerciales

Working Papers from HAL

Abstract: We document a Safe-Tail Paradox in banks' credit portfolios: retail borrowers classified as safest by scoring models are also the most exposed to artificial intelligence (AI)-related labor income risk. The paradox arises because AI exposure is positively correlated with borrower characteristics historically associated with low default risk (e.g., stable employment, high income), while AI exposure can weaken repayment capacity through displacement and wage compression. Credit risk therefore becomes concentrated in the safest segments of mortgage portfolios, precisely where regulatory capital buffers are thinnest. We design a borrower-level AI stress test and apply it to a synthetic portfolio calibrated to the French residential mortgage market. As AI adoption intensifies, capital requirements rise sixfold more in the safest class than in the riskiest, highlighting the need for AI-aware risk management.

Keywords: generative artificial intelligence; stress testing; retail credit risk; mortgage lending (search for similar items in EconPapers)
Date: 2026-04-23
References: Add references at CitEc
Citations:

There are no downloads for this item, see the EconPapers FAQ for hints about obtaining it.

Related works:
This item may be available elsewhere in EconPapers: Search for items with the same title.

Export reference: BibTeX RIS (EndNote, ProCite, RefMan) HTML/Text

Persistent link: https://EconPapers.repec.org/RePEc:hal:wpaper:hal-05688563

DOI: 10.2139/ssrn.6633858

Access Statistics for this paper

More papers in Working Papers from HAL
Bibliographic data for series maintained by CCSD ().