The rise of artificial intelligence (AI) in the banking sector is often seen as a promising advancement, aimed at optimizing the management of credit risks. However, a deeper analysis reveals a concerning paradox: this technology, designed to enhance financial security, could actually become a source of instability. This article examines how the integration of AI into the decision-making processes regarding bank loans can lead to unforeseen and potentially dangerous outcomes.
An ambivalent improvement tool
In the European banking landscape, AI is now indispensable. Modern algorithms are replacing traditional financial analysts, allowing for quick decisions on loan approvals. While this method has strengthened certain liquidity indicators, it is also correlated with an increase in non-performing loans, revealing underlying vulnerabilities within these institutions.
Regulators’ concerns
Prominent figures in the financial system, such as Christine Lagarde, President of the European Central Bank, have expressed growing concerns about the systemic risks generated by the increasing use of AI. In speeches delivered before the European Systemic Risk Board, she highlighted the urgent need to regulate these technologies to avoid a degradation in the quality of loans granted to borrowers.
An analysis of credit reports
To better understand the impact of AI on the banking system, we analyzed the quarterly financial reports of the main systemic banks in the euro area over a period extending from 2020 to 2025. By observing the frequency of AI-related terms, we developed an indicator of real usage of these systems in each institution, thereby enabling correlations with their risk levels.
A revealing paradox
The results of our study show that the increase in AI usage is accompanied by an improvement in liquidity management, but also by a deterioration in the quality of loans. Thus, an intensification of AI systems leads to a 7 percentage point increase in the liquidity coverage ratio, but also a rise of 0.35 percentage points in the non-performing loan ratio, representing nearly 14% of the average.
The limits of risk assessment
This situation raises questions about how AI assesses financial risks. The algorithms favor quantitative data, which is simple to process, but often neglect crucial qualitative elements, such as the unique context of a borrower. This trend can lead to the approval of loan applications that, while solid on paper, turn out to be risky upon closer examination.
The consequences for the real economy
The negative effects of this approach are not limited to the banking sector. The accumulation of non-performing loans linked to the use of AI could ultimately tighten the financing conditions of the private sector, with repercussions for aggregate demand. This poses real challenges for regulation, which must evolve to take these collective risks into account rather than individual ones.
The importance of transparency
To ensure effective regulation, it is essential that algorithms are not only efficient but also explainable. This is why the supervision of credit decisions must be rethought, incorporating systemic evaluation criteria. Such measures can help detect vulnerabilities that would otherwise be invisible in normal times.
Toward crisis scenario testing
It is crucial to test the responses of AI systems during economic crises through algorithmic stress tests. These tests would allow for the anticipation of problematic behaviors during times of stress, thereby helping to better prepare the financial system for potential shocks.
The indispensable role of humans
Finally, although automation can significantly enhance the efficiency of financial processes, it must never completely replace human judgment. The evaluation of complex financial situations requires a human approach that, combined with AI, could optimize risk without increasing fragility.
For more information on the implications of artificial intelligence in various fields, you can consult the following articles: Why every student must absolutely master AI?, Tips for detecting a retouched image, A look back at Eliza, the pioneer of conversational agents, celebrating its 60th anniversary, AI at the heart of recruitment, Innovative start-up and accessibility through AI.
