The AI Act, which is expected to enter into force this summer, defines two high-risk use cases for the financial sector: AI systems used to evaluate the creditworthiness of a person, and for risk assessments and pricing for life and health insurances of a person. The European Commission is planning to gather input from financial services stakeholders to get an overview of how and for which purposes AI applications are used in the financial sector. A targeted stakeholder consultation has been published (see below for more info), and workshops giving companies but also consumer associations the opportunity to present AI applications they are developing are being organised together with the European Supervisory Authorities and National Competent Authorities. But more concretely, how is AI being used in finance? And which opportunities and challenges does it bring?
Opportunities and use cases
The overall benefits of using AI/Machine Learning (ML) systems in the financial sector include increasing forecast accuracy, mitigating the risk of losses, automating processes, reducing costs, and increasing efficiency.
More specifically, the financial sector is increasingly harnessing the benefits of using AI systems for activities such as enhancing fraud detection and prevention. AI systems can analyse large amounts of (unstructured) data sets fast and detect anomalies or outliers that indicate fraudulent activity quicker than the human eye.
AI systems can also help to take more informed decisions by analysing data from a wide range of sources, such as market trends, economic indicators and customer behaviour. This will help institutions to have more information at hand, for instance, when deciding about investments or lending. Other use cases include algorithmic trading, customer services, or portfolio management.
What are the challenges?
However, AI comes with certain risks. Some of them relate to respecting data protection regulations. Others concern the AI system itself. The trustworthiness of an AI system can be difficult to determine if the quality of data is not sufficiently clear. A sensitive issue related to this is algorithmic bias, which can lead to discrimination. An AI model can reproduce or even amplify biases and discriminatory patterns that were mirrored in the data used to train the model. This is also why ‘explainability’ is a pivotal challenge for AI systems – the ability to explain why a certain decision was taken and which parameters were used. For example, why a person was (not) granted a loan.
What are the next steps?
The outreach of Commission Directorate General for Financial Stability, Financial Services and Capital Markets Union (DG FISMA) is twofold
- First, the targeted consultation which is structured in three parts. It asks what the general benefits and challenges linked to AI in finance are and if the existing financial rules are affected. It also deals with sector-specific use cases (banking, market infrastructure, securities, insurance and pensions, asset management). The last part includes questions on the requirements linked to the AI Act and understanding the needs of the financial sector for its implementation
- The AI workshops, meanwhile, provide a more informal platform to showcase recent AI finance tools that are being used or developed. The workshops aim to provide further insight into how AI in finance looks in practice, and offer an opportunity for the AI finance community to exchange ideas, inspire each other and hopefully foster innovative thinking
Related links
Details
- Publication date
- 19 June 2024
- Author
- Directorate-General for Financial Stability, Financial Services and Capital Markets Union