The European Association of Co-operative Banks (EACB) is happy to contribute to the discussion on the Artificial Intelligence (AI) legislative proposal.
The EACB recognises that the AI proposal is the Commission’s first ever legal framework on the matter, which addresses the risks of AI and aims to position Europe to play a leading role globally. It should be recognised that this is a risky bet. If European values were not ultimately adopted on an international scale, European companies would be at a disadvantage compared to non-European players active in less restrictive regulatory environments.
We believe that the European Commission, the European Parliament and the Council should remain vigilant to ensure that European players are not unduly constrained in their prospect of developing innovative AI solutions compared to international competitors.
The EACB welcomes the Commission’s risk-based approach as basis for a proportionate legal text. We appreciate the technology-neutral and future-proof definition of AI, recognising that AI is a “fast evolving family of technologies” that is constantly developing. Nevertheless, combining the definition of artificial intelligence system together with the techniques and approaches of Annex I of the proposal, we observe that the scope of the Regulation is becoming quite wide as it also includes rule-based approaches.
We believe it is of paramount importance to make sure that the AI proposal will not add new and burdensome requirements for the banking sector and create conflicts and overlaps with existing rules: e.g., sector-specific regulation (CRD, CRR).
Generally, some provisions of the Regulation contain somewhat vague wording, e.g., the definitions provided for “remote biometric identification system” and “user”. Moreover, we believe that the definition of ‘developer’ and ‘end user’ are missing from the legal text. These points should be further clarified in order to guarantee legal certainty for providers, developers and users of AI systems.
Regarding regulatory sandboxes, EACB members believe they are useful for the development of AI. However, their objectives and entry criteria should be clear and made public in order to ensure a high degree of transparency and a level playing field in the entry process.
Finally, we wonder how AI systems can be prevented from being biased. This requirement is not realistic as it cannot be guaranteed that datasets will be fully correct or complete. We believe that errors should be minimised and that the training, validation and testing of the AI system should be as complete as possible.