欧洲中央银行关于简化人工智能协调规则实施的法规提案的意见

OPINION OF THE EUROPEAN CENTRAL BANK of 13 March 2026 on a proposal for a regulation as regards the simplification of the implementation of harmonised Introduction and legal basis On 19 November 2025, the European Commission published a proposal for a regulation of the EuropeanParliament and of the Council amending Regulations (EU) 2024/1689 and (EU) 2018/1139 as regards the The European Central Bank (ECB) has decided to deliver an own initiative opinion on the proposedregulation. The ECB’s competence to deliver an opinion is based on Articles 127(4) and 282(5) of the Treaty 1.General observations 1.1In 2021 the ECB received a request from the Council of the European Union for, and adopted, anopinion on the original proposal for a regulation of the European Parliament and of the Council laying 1.2TheECB welcomes the objective of the proposed regulation to promote innovation andcompetitiveness within the internal market by simplifying and streamlining the implementation ofRegulation (EU) 2024/1689of the European Parliament and of the Council3(hereinafter the ‘AI Act’). The ECB acknowledges the importance of establishing an accessible regulatory framework specific particular, as credit institutions increasingly rely on both high-risk and general-purpose AI systems,a streamlined framework is essential to ensure compliance with the AI Act and Union banking law. 1.3From a prudential supervisory perspective, the establishment of well-designed innovation pathwaysallow credit institutions to safely experiment with AI solutions in controlled environments, while 1.4The ECB understands that the proposed regulation does not impose additional obligations on theECB in its capacity as prudential supervisor of credit institutions, nor does the ECB need to apply theAI Act in its supervision. Nonetheless, due to the ECB’s supervision of credit institutions’ information and communication technology (ICT) risk, and the fact that certain requirements under the AI Act 2.Institutional competences and supervisory powers 2.1Under Article 127(6) of the Treaty and Council Regulation (EU) No 1024/20136(hereinafter the ‘SSMRegulation’), the Council has conferred on the ECB specific tasks concerning policies relating to theprudential supervision of credit institutions, with a view to contributing, inter alia, to the safety andsoundness of credit institutions and the stability of the financial system within the Union and eachMember State7. The ECB’s mandate is limited to the prudential supervision of credit institutions. The 2.2Notwithstanding the ECB’s restricted mandate in this field, the AI Act provides that national marketsurveillanceauthorities,supervising regulated credit institutions participating in the Single in the course of their market surveillance activities that may be of potential interest for the ECB’s 3.Exchange of supervisory information 3.1Credit institutions that are providers or deployers of high-risk systems are deemed to comply withcertain provisions of the AI Act when they comply with similar obligations under Union financial services law13. The ECB welcomes this principle, which prevents credit institutions from being facedwith double reporting and potential conflicting supervisory views. However, as indicated in paragraph2.1, the ECB does not have a market surveillance mandate under the AI Act and, in general, banking 3.2The current framework primarily envisages the ECB as a recipient of information from market surveillance authorities14. In order to achieve the objectives of the AI Act, including the protection offundamental rights,and ensure effective and coordinated market surveillance and prudentialsupervision, the proposed regulation should include an explicit legal basis allowing the ECB, on aneed-to-know basis and subject to professional secrecy requirements, to share relevant prudential supervisory information with competent national market surveillance authorities supervising creditinstitutions which are regulated under the CRD and which fall under the ECB’s supervision. A similar 4.Scope and classification of models and use cases 4.1It is important for market participants, in particular credit institutions that use AI for credit scoring, toknow which techniques fall under the definition of AI system17and which do not, as well as which 4.2Further clarification would be appropriate, particularly by expressly excluding generalised linearmodels (e.g. linear or logistic regressions) from the scope of the AI Act’s definition of high-risk AIsystems, when used for credit scoring18, due to their high level of explainability and transparency19. 4.3In this regard, the ECB notes that generalised linear models, including linear and logistic regression, are inherently interpretable and transparent22. Their operation can be fully explained through a limitedand stable set of parameters, whose influence on outcomes can be assessed using well-established statistical tools. Consequently, such