On the one hand, machine learning has enabled new connections and problem solutions to be found in the form of practically implementable algorithms that without these technologies would have remained concealed. On the other hand, it is often not possible to explain in detail how some of these algorithms are discovered by the machine. In isolated cases, these algorithms result in errors that are difficult to explain. There are also examples of such algorithms that – unintentionally – lead to systematic discrimination.
Owing to the abundance of data, the financial market offers a promising field for the use of AI. Based on surveys at selected banks, FINMA compiled an initial review of the areas of use for AI applications in the year under review. These include
FINMA is planning further work and analyses in 2022 concerning the use of these technologies in areas relevant to supervisory law. It will concentrate on data management, governance and the control environment of the application.
A detailed survey of one hundred Swiss insurance companies conducted by FINMA showed that AI is already widely used in this sector. The insurers use it mainly in customer interactions, as well as in claims processing and distribution. AI is also used for pricing applications. In relation to governance, the insurers have begun to set up committees to professionalise and further develop their AI-specific processes.
FINMA has also developed initial applications based on AI that are used to support data-driven supervision. These tools have been used for a number of years and include automated data evaluation for analysing irregularities. Such applications can be found in the Asset Management division with respect to certain investment vehicles and in the Banks division in connection with selected financial institutions.
(From the Annual Report 2021)