The European Banking Federation (EBF) welcomes the opportunity to comment on the European Commission’s proposal for a Regulation laying down harmonized rules on Artificial Intelligence (AI Act).
The Act aims to develop an ecosystem of trust for AI in the EU, which encourages investment, aims to strengthen competitiveness and ensures that AI systems respect fundamental rights and EU values. It is essential for these rules to strike a balance between regulating high-risk AI use cases and supporting innovation, including proportionate requirements, no unjustified barriers or restrictions, and no duplication.
Coherence with existing, sectoral regulation and horizontal regulation, notably the GDPR, is also crucial. The Regulation should not lead to a duplication of requirements. The banking and financial services sector is already subject to strong sectoral regulation and supervision, which ensures consumer protection, risk management and financial stability in all services provided to customers, regardless of whether those applications or services involve the use of technologies such as AI, including in the case of creditworthiness assessment.
On supervision, while we welcome the designation of financial authorities as Competent Authorities for credit institutions’ activities, the supervision under the Regulation risks creating an unlevel playing field across different countries and industries if there is no consistency in supervisory expectations and practices among different national competent authorities. These divergences could occur when the same high-risk AI application, such as creditworthiness assessment and credit scoring, is concerned since different entities might be supervised by different market surveillance authorities. In order to ensure a level playing field for all industries in the application of the Regulation, the principle of ‘same activity, same risks, same rules’, must be taken into account, while ensuring a well-coordinated and harmonized supervisory landscape for all market participants providing or using high-risk AI systems and maintaining a high level of consumer protection to ensure that consumers have confidence and trust in the use of AI.
The EBF highlights the following recommendations:
A more targeted approach for the definition of AI is needed to distinguish between different systems and the scope of applicability. Given that the proposed definition will be the first definition of AI included in an EU Regulation and would be the reference for other potential rules referring to these technologies, the definition must be fit for purpose.
More guidance to clearly identify when the use of AI systems will be considered high risk and therefore subject to the requirements set out in this Regulation. This is particularly important in light of firms innovating and developing new use cases that may not clearly fit into the categories set out in Annex III.
Clarifying the scope of the creditworthiness assessment use case in the text - only those systems used to evaluate the access to credit should fall in the scope; AI applications used in the wider credit process should be excluded from the scope.
While the designation of financial authorities as Competent Authorities for credit institutions’ activities is welcome, ensuring i) coherence and a clear delineation with the proposal for a revision of the Consumer Credit Directive[1] (CCD) and ii) alignment with the EBA guidelines on loan origination, is needed.
To ensure a level playing field, guidelines or criteria are needed for each authority to follow and comply with for a harmonized interpretation and application of the AI Regulation for all high-risk use cases in order to avoid the risks of and that certain actors are subject to more stringent requirements in one member state than another.
The missions of the new European Artificial Intelligence Board must be better defined in light of the challenges of international competitiveness, innovation, and data processing (personal or industrial). The role of existing authorities and how they fit within this new constellation must be specified.
Including provisions on promoting measures that that support customer education and awareness raising on the Regulation and the use of AI, its real capabilities and its benefits for their users in order to address doubts at the member state level.