BaFin - Navigation & Service

Go to:

Symbolfoto @ Metamorworks/stock.adobe.com

Topic Fintechs Big data, artificial intelligence and machine learning

Many people would agree that big data (BD) and automated data processing using artificial intelligence (AI) and/or machine learning (ML) are the key to digital transformation in almost every industry. In the financial sector, too, there are application scenarios that would be barely conceivable without BDAI/ML.

In simple terms, BaFin defines AI as the combination of big data, computing resources and machine learning. However, this definition does not make a universally applicable distinction between BDAI/ML methods and other methods. In fact, when it comes to a specific method to be examined, supervisory practice, inspection techniques and inspection intensity need to take into account which BDAI/ML characteristics are present and the extent to which these characteristics are pronounced – there are thus no clear-cut lines between the methods and there might be some overlap.

The characteristics can be grouped into three dimensions of the BDAI/ML scenario:

(1) The methodology and data basis describe the complexity and thus the model risk associated with the BDAI/ML process.
(2) The use of the output relates to the significance of the process within a supervised entity’s risk management.
(3) The third dimension relates to the general distinction between in-house development and outsourcing and the underlying IT infrastructure.

(See BaFin/Bundesbank, Consultation 11/2021 – Consultation paper: Machine learning in risk models – Characteristics and supervisory priorities https://www.bundesbank.de/en/homepage/machine-learning-in-risk-models-characteristics-and-supervisory-priorities-793670)

In the banking and insurance sectors, BDAI/ML is typically used for anti-money laundering and fraud detection and for back office processes (such as black box processing in the insurance sector). BDAI/ML is also used for scoring, pricing (incl. telematics rates) and for sales support. In the area of risk management, BDAI/ML is increasingly being used to prepare data and validate risk models. In the securities industry, which is already characterised by a high level of automation, BDAI/ML is primarily used to enhance existing processes in trading, advice, risk management and compliance.

In 2021, BaFin defined general principles for the use of algorithms, with a focus on algorithms in the area of BDAI/ML (https://www.bafin.de/dok/16178870). In doing so, BaFin set out key general principles (e.g. clear management responsibility) and specific principles for the development phase and application phase of BDAI/ML applications. For example, data strategy, data governance and the implementation of appropriate validation processes must be taken into account in the development phase, while in-depth approval and feedback processes must be carried out during the application phase.

Risk associated with BDAI/ML methods

BDAI/ML method performance can prompt users to blindly trust the data

The data on which BDAI/ML methods are based should be viewed as a starting point and a success factor. Unstructured data can now be exploited by and for BDAI/ML methods. In addition, BDAI/ML methods allow for calculations that factor in a large number of determinants, which in turn may lead to overfitting. When large volumes of data are used, the quality of this data must be continuously ensured. This not only applies to model development and validation but also applies to model application. (See BaFin/Bundesbank, Consultation 11/2021 – Consultation paper: Machine learning in risk models – Characteristics and supervisory priorities)

Focus on explainability

As the complexity and the number of dimensions of a model’s hypothesis space increase, it becomes more difficult to describe the functional relationship between input and output either verbally or with mathematical formulas. The calculations are then more difficult to understand for those modelling, using, validating and supervising the model in question. It can also be more complicated to check the validity of the model’s output. User acceptance may suffer, too. Although such a “black box” characteristic may be justified, e.g. if this results in higher predictive performance, it can lead to potentially greater model risk. Explainable AI (XAI) methods have been developed to address this risk appropriately. But even if XAI methods seem highly promising from a supervisory perspective as a means to mitigate the impact of this “black box” characteristic, these XAI methods are also models with assumptions and weaknesses, and, in many cases, they are still being tested. (See BaFin/Bundesbank, Consultation 11/2021 – Consultation paper: Machine learning in risk models – Characteristics and supervisory priorities)

Adaptivity: model changes are more difficult to identify

In the banking sector, institutions are required to report changes to Pillar 1 models to supervisory authorities, and, in some cases, they may implement these changes only after they have been approved. There is no clear-cut distinction between regular model maintenance and model changes, especially since the meaning of the term “model change” also depends on the relevant supervisory context. The flexibility and, in some cases, high-frequency adaptivity of BDAI/ML processes make it more difficult to draw a clear line between adjustments and changes; such a clear distinction, however, is indispensable for supervisory purposes.
As a general rule, the need for high-frequency adaptivity should be thoroughly justified. From a supervisory perspective, it is crucial to adapt the training cycle to the specific use case and to provide the necessary justification in order to achieve a balance between ensuring that the data is up-to-date and ensuring that models can be explained and validated.
(See BaFin/Bundesbank, Consultation 11/2021 – Consultation paper: Machine learning in risk models – Characteristics and supervisory priorities)

Additional information

More on this topic

Did you find this article helpful?

We appreciate your feedback

Your feedback helps us to continuously improve the website and to keep it up to date. If you have any questions and would like us to contact you, please use our contact form. Please send any disclosures about actual or suspected violations of supervisory provisions to our contact point for whistleblowers.

We appreciate your feedback

* Mandatory field