The financial industry fully agrees on the potential of AI and the need to have a strong leadership in the adoption and development of AI in the EU to stay competitive at global level.
For financial services, AI approaches can enable the customization of products to clients’ preferences, needs and expectations. Tasks previously done manually can be performed more efficiently and with higher accuracy, improving internal processes by reducing false positives, or enhancing the claims management processes, including improved detection of fraudulent claims. AI can help insurers to refine existing actuarial models or processes, enhancing the availability and innovativeness of insurance for EU customers.
There is also a huge potential for innovation and broader benefits. A particularly promising area would be related to risk management, where AI and data analytics can contribute to enhancing financial institutions risk analysis capabilities, both helping more people get access to financing, removing unintentional bias in existing processes, while also improving the sustainability and resilience of the financial system.
The Artificial Intelligence Act aims to ensure the safety of AI by safeguarding the EU’s fundamental rights. While AI systems can vary greatly from one another, it is important that this regulation addresses specific higher risks uses of AI and aims to mitigate these risks for consumers and society at large. Therefore we welcome the risk-based approach followed by the proposed AI Regulation as we think it could support trustworthy AI use cases with embedded EU values.
However, a balance must be struck between ensuring security and raising awareness on potential risks related to AI and fulfilling it innovative potential The proposed text must be enhanced to avoid regulatory uncertainties leading to unexpected regulatory burdens and, therefore, hold back the development and adoption of AI in the EU.
There are some uncertainties in the Commission’s text that, depending on the final interpretation, could significantly increase the final impact of the Regulation. Those uncertainties can be grouped into three categories: the scope of the Regulation, the governance framework, and the requirements imposed. In its Paper the EFR gives recommendations on these three categories.
A European AI Regulation that promotes the use and adoption of trustworthy AI embedding EU values should be proportionate and clear. This would avoid creating legal uncertainty or increasing (unnecessary) compliance burden.