06/09/2026
The ECB will incorporate the use of AI into its banking risk and solvency assessments
The impact of technology will be incorporated into SREP exercises as a cross-cutting risk
Artificial intelligence (AI) is making a major inroads into banking supervision. The European Central Bank (ECB) will incorporate the use of this technology and the risks associated with it into its analysis of institutions as part of its annual Supervisory Review and Evaluation Process (SREP), at a time when the financial sector is pulling out all the stops to maximise the benefits of adopting AI tools to improve lending and customer service, detect fraud or, quite simply, optimise results by achieving efficiencies through process automation.
The technology has been under supervisory scrutiny for some time, but specific scrutiny of AI usage will now form part of the risk analysis in SREP exercises, following the implementation of the European Artificial Intelligence Act (AI Act), adopted in 2024, which this year will mark a turning point in its phased implementation through to 2028.
“For credit institutions, the most important deadline is 2 August 2026, the date from which AI systems used in areas considered high-risk must fully comply with the AI Act’s requirements regarding governance, data quality, risk management, human oversight, transparency, documentation and monitoring”, explains Carla Azorí, Senior Regulatory Affairs Manager at Accuracy. Systems classified as high-risk include those used to assess customers’ creditworthiness, determine their credit rating or set actuarial pricing for insurance policies.
Azorí highlights that the ECB incorporates the impact of AI “within the 2026 SREP guidelines” not as a standalone risk but “across the board, primarily through the areas of digitalisation, governance, operational resilience, ICT risks, data management and business models”. In practice, it will assess the impact across all areas, as is the case with ESG (environmental, social and governance) risks.
“AI can significantly improve the efficiency, accuracy and predictive power of credit risk management, but its use requires a robust framework for governance, data protection, human oversight, bias control, transparency and regulatory compliance,” adds the Accuracy executive.
The advantages of AI are clear: it helps to improve customer service through greater personalisation of services and offerings, optimises credit assessment and enables faster responses via the channel of the user’s choice. Furthermore, it makes it easier to anticipate customers’ needs, automate processes whilst minimising operational errors, and equip the bank’s staff and professionals with better tools and resources for their tasks.
Operational resilience
As part of the SREP exercise, the supervisor must assess the extent to which new technologies have been adopted and their impact on strategy, revenue generation and cost control – but not only that. The exercise requires the institution’s board of directors to possess sufficient knowledge and expertise regarding emerging risks, explicitly including ICT (information and communications technology), which encompasses those linked to AI. It also forms part of the assessment of operational resilience – including connections with third-party providers – and of the quality of data, ICT systems and reporting. It is also central to the analysis of operational resilience.
“In an increasingly digital environment, an organisation’s strength will depend not only on its ability to innovate, but also on its ability to manage the risks that such innovation generates,” adds Azorí.
Perhaps the main threat posed by AI to the industry lies in the area of cybersecurity, as it enables criminals to automate large-scale attacks. Mythos’s advanced AI model has put the sector and the regulator on alert. The ECB has required the major eurozone institutions to draw up and submit a comprehensive action plan to address the major cybersecurity risks posed by Anthropic’s programme by 31 October.
Source: El Economista
