Recommendations when using supervised machine learning

July 10th 2019, the Danish FSA published a set of recommendations for financial institutions on the use of supervised machine learning (“God praksis ved brug af superviseret machine learning"). The paper is based on experience gained by the Danish FSA, including from the regulatory sandbox FT Lab and from the cooperation with European supervisory authorities.

One of the key learnings at the Danish FSA has been that when using supervised machine learning financial institutions must be aware of the new risks their business models are exposed to. The risk picture might change when using the new technology and the Danish FSA would expect financial institutions to ensure that their existing governance structures can handle the use of machine learning.

”Financial institutions should fully realize their purpose of using machine learning. The purpose must be clear to the organization and should support that subsequent choices are handled properly,” says Director for payment service, fintech and governance Tobias Thygesen of the Danish FSA.