Desafíos actuales de la Inteligencia Artificial

74 Desafíos actuales de la Inteligencia Artificial 1. INTRODUCTION The development of the EU digital economy has been at the centre of several policy strategies in the last decades (Mariniello, 2022). In this respect, a wide range of technological developments have caught the attention of EU policy and lawmakers; however, there is one topic that has received further interest: data. Currently, the main policy document dealing with it, from which different regulatory actions emerge, is the EU 2020 Data Strategy (Euro- pean Commission, 2020). At the same time, the relevance of artificial intelligence (AI) developments has also taken ground in the vision of the EU digital economy. On the one hand, the EU is actively promoting the concept of data spaces—shared ecosystems designed to facilitate secure and efficient data sharing and collaboration un- der its EU 2020 Data Strategy, alongside the development of different regulatory in- struments to ensure the free flow of data across the EU (Chomczyk Penedo, 2024). To facilitate data-driven innovation, these data spaces would enable the sharing of (person- al) data between different stakeholders, as long as compliance with EU data protection laws is followed (Curry, Scerri & Tuikka, 2022). By developing these, the EU intends to consolidate a common single market for data across its Member States. However, while these data spaces would enable the sharing of data, they also introduce restrictions on how data can flow between different parties, for example one of the regulatory proposals that integrate the European Financial Data Space, the Financial Data Access Regulation (FiDAR proposal). 1 On the other hand, to tackle a wide range of potential harms and risks that the use of AI can produce, the EU has adopted the AI Act, which represents a groundbreaking regulatory step. From the wide range of potential issues that could compromise fundamental rights, the existence of biases within these systems has demonstrated a serious threat (Laupman, Schip- pers & Papaléo Gagliardi, 2022). In this respect, certain personal data categories, such as racial or ethnic origin or trade union membership, have long been excluded as characteristics to consider in decision-mak- ing, given their potential to expose individuals to further discrimination (Kelly et al, 2022). However, research in algorithmic discrimination field has shown that the processing of these categories can help to overcome existing biases in AI systems (Hoffmann et al ., 2022). As such, Article 10(5) of the AI Act specifically addresses the use of special categories of personal data to detect and correct biases. 1 Proposal for a REGULATION OF THE EUROPEAN PARLIAMENT AND OF THE COUNCIL on a fra- mework for Financial Data Access and amending Regulations (EU) No 1093/2010, (EU) No 1094/2010, (EU) No 1095/2010 and (EU) 2022/2554 COM/2023/360 final. This proposal should be read alongside the update to the current open banking framework under the Proposal for a REGULATION OF THE EUROPEAN PAR- LIAMENT AND OF THE COUNCIL on payment services in the internal market and amending Regulation (EU) No 1093/2010 COM/2023/367 final.