Faced with the growing adoption of AI in Swiss banks, FINMA highlights the risks associated with its use and calls for robust governance and continuous monitoring.
The Federal Financial Market Supervisory Authority (FINMA) recently published a communication detailing its expectations regarding the management of risks linked to the growing use of artificial intelligence in Swiss financial institutions.
In FINMA’s view, AI presents considerable opportunities for financial institutions, particularly in terms of operational efficiency, personalization of services and fraud detection. The financial watchdog emphasizes, however, that the use of these technologies is accompanied by risks that are often complex to assess. Among the risks identified by FINMA are the lack of accuracy of responses generated by large language models, algorithmic biases, data security and quality, cyberattacks, as well as dependence on third-party providers.
FINMA indicates that it has observed various shortcomings in terms of governance and risk management linked to AI within Swiss banks. For example, often decentralized development makes it difficult to implement standards and assign clear responsibilities. The absence of systematic criteria to identify applications requiring particular attention is also pointed out.
Banks are therefore invited to strengthen governance aimed at ensuring adequate management of these risks, with well-defined responsibilities and transparent documentation. FINMA also recommends a strict separation between the development of AI applications and their independent verification by qualified personnel in the field. The authority further advocates regular testing and constant monitoring mechanisms to ensure the reliability of AI applications. Furthermore, clear documentation and precise criteria for evaluating the explainability of models are necessary.
News-Swiss
