Technology Towards 2035: The Future of Lending and Supervision

OCCASIONAL PAPER Technology Towards 2035 The Future of Lending and Supervision

In Brief The digitalizing world undoubtedly impacts the financial sector. For this exploration, the AFM started from the technological possibilities that the digitalizing world may bring in the coming years. These have been translated into a future vision for the credit sector, specifically for mortgage and consumer credit. With this future vision, we aim to sharpen our understanding of the long-term opportunities and risks of possible developments in the sector and their impact on AFM supervision. Additionally, we want to provide a foundation for targeted discussions with stakeholders, to reduce the risks of digitalization in the future while simultaneously seizing the opportunities to improve products and services for consumers.

Authors: Mark Teunissen, Bart Zwartjes, Tomas Bolscher, Jules van de Ree and Joshua Meijer Publication date: June 2025

OCCASIONAL PAPER Table of Contents Introduction 3 Management Summary 4

1. Technological Trends 7 1.1 There is more and more data available 7 1.2 Data ecosystems ensure data storage and access 7 1.3 Data analysis and especially AI are taking off enormously 8 1.4 Protection of digital security becomes top priority 8 1.5 Smart contracts in the form of code can mark agreements 9 1.6 Digital identity provides a new way of identification 9
2. Impact on Lending 10 2.1 Process of lending 10 2.2 New propositions, new players 15 2.3 New dependencies 16
3. Issues for Supervision and the Sector 17 3.1 Instant credit 17 3.2 Access to lending 18 3.3 Personalization of credit 19 3.4 More complex and interconnected ecosystem around lending 20 Contents

OCCASIONAL PAPER Technology Towards 2035 3 Introduction 1 Think, for example, of the second Consumer Credit Directive (CCD II) and the AI Act. Although this exploration points out important restrictions on the use of data in lending resulting from this new regulation in several places, we do not intend to provide an exhaustive analysis of the regulation that may influence the possible developments described. In the AFM Strategy 2023-2026, we identify three long-term and fundamental trends that influence the functioning of financial markets and will have an impact on the course and supervisory approach of the AFM. One of these trends is digitalization. 2026 is near. The world and our supervision are in motion. What impact will digitalization have in the long term on an important supervisory theme – consumer lending? In 2023, we launched an exploration of the insurance sector. This report focuses on the lending sector and more specifically: on the financial products mortgage and consumer credit. Once again, the impact of digitalization is central, looking ten years ahead – to 2035.

Scope and Approach In this exploration, we have not been limited by existing and upcoming legislation. We started from the technological possibilities that the digitalizing world may bring in the coming years. From this analysis, we then translated this to the lending sector and to the opportunities, risks, and dilemmas that may arise from it – both for the sector and for us as supervisors. A thought experiment on the possibilities and implications of digitalization, translated to mortgage and consumer credit, without directly anticipating all conceivable future measures and regulation that are currently known or even unknown.1 The insights from this exploration come from literature research and from conversations with stakeholders within and outside the sector.

Objective With this exploration, looking at the world up to 2035, we aim to: • Map the opportunities and additional risks of digitalization for a specific sub-market, including the dilemmas that arise from it. • Kick-start a dialogue with the sector, supervisors, regulators, policymakers, and other stakeholders. With this exploration as the foundation for this conversation.

Reader's Guide After the management summary, we discuss the most important technological trends (Chapter 1), followed by the opportunities and possibilities they offer (Chapter 2). The report ends with the issues for supervision and the sector (Chapter 3).

OCCASIONAL PAPER Technology Towards 2035 4 Management Summary Digitalization will have an impact on consumer lending in the coming years. The way credit is offered, but also the forms of credit and the parties involved in lending, may change under the influence of technology and digitalization.

Technological Trends Important technological trends are the increasingly large amount of data, accompanied by the surge in data analysis and especially artificial intelligence (AI). Data is the foundation on which the digital system is built, and data sources continue to expand. Opportunities for lenders to access data sources from other organizations also arise due to open finance legislative initiatives in the European Union (EU). AI can be applied in many processes of the lending chain. Access to more and more data points of an individual and the increased possibilities of data analysis make it possible to segment customer groups extensively. Protection of digital security becomes essential – cybercrime permeates all levels of the digitalizing world.

Impact on Lending More than 2025, 2035 will be the time of 'one click away': everything can be everywhere and direct, interactive, and digital. Lenders will strive as much as possible for 'one click away'. The credit process will be further digitized, and become more personalized and interactive. For consumers, this will lower barriers and increase ease of use. This gives positive effects, such as early signaling of payment problems and low-threshold payment solutions. But it also gives negative effects, such as debt habituation. By using source data combined with a digital identification wallet, the credit acceptance process can be further digitized. The consumer no longer submits documents with the credit application. A third party – or the lender itself – retrieves all source data from source parties (such as the UWV or the Tax Authority) with the customer's consent and shares it with the lender. Lenders can use an AI-driven document management application. The application builds a credit file in line with European law and regulation. This makes the process potentially less error-prone and more cost-efficient. For example, the software issues warnings if data is incorrect or inconsistent, or if documentation is missing. Improvements in technology around digital ID verification and the use of source data can also contribute to secure lending with less consumer fraud. At the same time, there are more risks of spoofing, phishing, and other forms of hacking. Also due to the mentioned European data market developments, chain parties increasingly have more and different types of data to accurately, finely, and quickly determine the consumer's creditworthiness. Using payment transaction data, there is visibility into spending patterns, payment behavior, and repayment history, and thus into the current net disposable income of customers. Potentially, lenders can also use alternative data sources to map the creditworthiness, and thus the risk profile of someone, in detail. Technically, it is possible to include the model-predicted future income perspective of a consumer in the credit assessment – something that is already common in business lending.

OCCASIONAL PAPER Technology Towards 2035 5 Through deeper insight into the customer's personal financial situation, the credit offer can eventually be further personalized. This could manifest itself in personalized pricing. Customer-specific risks could translate into credit pricing, personalized marketing and advice, and personalized product propositions, for example for specific niche situations. This fits into society in 2035, where consumers expect personalized solutions and services for their individual requirements and needs. AI can play an increasingly larger role in customer contact. Initially to support the advisor, but possibly also in a more comprehensive role. Customers are then guided through the process of lending – for example via a mortgage advice portal. The customer essentially completes the customer journey autonomously and digitally, but the advisor is on standby at every phase. The advisor is available if the customer indicates a need for personal advice. Technology offers possibilities to subsequently follow the customer more targeted in the period after the credit is closed, so that one can better anticipate changing circumstances and major life events. Technology can promote the development of innovative credit propositions. It also seems likely that embedded lending – consumer credit integrated into non-financial products using digital solutions – will take off more.

Issues for Supervision and the Sector The AFM, as a conduct supervisor for the lending sector, will also in 2035 supervise the principle of 'customer interest central', with sufficient (duty of care) safeguards for the consumer. This also requires supervision that must anticipate the further digitalizing world and the changing lending context. Besides the opportunities that technology offers for better consumer lending described above, risks for consumers also arise. For the AFM, the interest of adequate management of these risks is paramount. Therefore, we will be alert to these risks and the extent to which they can harm customer interest. However, we continue to keep an eye on the opportunities that digitalization offers for improving the quality of and access to financial services for consumers. This requires a balance between giving room for utilizing these opportunities and mitigating possible risks. Against this background, the AFM will explicitly pay attention in the coming years to a number of issues that emerge from this exploration. A first issue concerns instant credit. Lowering barriers to access to lending offers opportunities for the consumer. For example, greater ease of use and greater self-reliance. At the same time, a gradual cultural shift may occur from 'credit is the exception' to 'credit is the norm'. Immediate access in a digital environment to new credit propositions increases the chance of impulse purchases financed with credit, and the chance of debt stacking. And through the use of digital applications, the risk of fraud by malicious parties increases, especially among customer groups that are less digitally familiar. A second issue is how the consumer can be protected against risks of exclusion from lending, without this coming at the expense of opportunities for greater accessibility. Although digitalization can lead to underserved groups getting more and/or better access to lending, it cannot be ruled out that a more data-driven credit assessment will lead to the exclusion of customers with an atypical risk profile. Another risk is that digitalization will hinder access to credit for people who are less digitally skilled. Or that people who want to do everything offline will eventually receive less favorable conditions (such as an interest rate surcharge) for service that is essentially comparable. A third issue focuses on dealing with the risks resulting from the personalization of the credit offer. Digitalization can lead to the personalization of lending and thereby addresses the needs of consumers. But this is inextricably linked with dilemmas regarding privacy and ethics. There may be consumers willing to share more personal data with the lender with the goal of receiving a more favorable offer compared to customers who do not want to. This can lead to risks of division and unequal treatment among consumers. The trend towards personalization may also make credit products more complex. And there is a risk that opaque credit markets will emerge; markets with non-transparent, highly personalized pricing and terms. Whereby the risk cannot be ruled out that the customer ultimately pays too much or receives an offer that is unattractive to the customer. A fourth issue is the growing challenge of being able to adequately (supervise) a increasingly complex and more interconnected ecosystem around lending. In the value chain of lending, increasingly diverse and heterogeneous players will operate, partly also established in different countries within and outside the EU. The activities, roles, and responsibilities of traditional lenders are shifting to other actors. The chain thus becomes more complex, diffuse, and interconnected. As a result, behavioral and operational risks may arise outside the view of supervision. Many of these parties also fall outside the mandate of the supervisor. Due to digitalization, the point of reference for supervision of lending also changes. Supervision will, for example, have to focus more on testing whether algorithms comply. This places different demands on supervision. And finally: some of the described technological trends also necessitate thinking about how these trends relate to existing laws and regulations regarding conduct supervision and privacy. From the conversations held during this exploration, the image emerges that the current progress in digitalization and standardization and the extent to which innovation is embraced in the credit chain seem more limited than what is technically possible. For example, it seems that a possible tension exists between current regulation and the possibilities that technology and digitalization already offer in several parts of the lending process. This raises the question of to what extent some existing rules are still future-proof. This tension is particularly evident in the loan norms; tension that is the result of the possibilities of technology and the possible use of source and payment data. Furthermore, this exploration outlines a number of privacy issues which, incidentally, primarily fall under the jurisdiction of the Dutch Data Protection Authority (Autoriteit Persoonsgegevens). We will have to align our supervision as much as possible with the developments mentioned in the coming years. In addition, we call on the sector, but also other involved supervisors, regulators, and policymakers to, from their role and responsibility, think about the possible future perspective outlined in this exploration and the issues that arise from it. This is to reduce the risks of digitalization in the future while simultaneously seizing the opportunities to improve products and services for consumers. The AFM is happy to continue this dialogue.

OCCASIONAL PAPER Technology Towards 2035 7

1. Technological Trends 1.1 There is more and more data available Data is the foundation on which the digital system is built, and data sources continue to expand. Every action and interaction in the digital world is recorded in the form of data. This data arises from interactions with, or actions of, the customer in the organization's digital environment (internal data), or from cooperation agreements or publicly available data sources (external data). Until recently, organizations mainly had access to data they had built up themselves. Partly due to legislative initiatives in the EU, such as PSD3 and FIDA in the context of open finance, opportunities arise to access data sources from other organizations – besides the growing public data sources. However, existing legislation, such as the GDPR and requirements regarding purpose limitation and data minimization, must be met. In a digitalizing world, almost everything can be recorded in data. For lending, it is important to assess the risk of financial problems and payment arrears as accurately as possible. Because more and more organizations are becoming digital, more relevant sources are becoming available to make such an assessment increasingly accurate. Here, it is essential to determine where relevant data is available, what data is exactly relevant, and how access to it can be obtained. The ability to unlock all relevant data also offers ease of use and reduces processing times for potential customers. Continuously determining the relevance, quality, and reliability of data will be a challenge for the future. Data can become outdated or prove to be incorrect. The highest data quality and reliability are necessary to guarantee the quality of the models. Incorrect outcomes can have major consequences. Data comes in many forms, and the value of data is not limited to text and numbers alone. Advanced techniques make it possible to translate audio to text, analyze photos and videos at the pixel level, and translate biometric input into an identifiable person. A trend break may occur from data maximization to data minimization. For now, there is a need to collect data for own analyses or as a basis for new propositions. It is not unthinkable that in the near future, also due to the importance of privacy, work will be done with more aggregated data. Think, for example, of the concept of zero-knowledge proof, where an organization does not need a birth date or identification document for an age check, but asks an institution (for example, a government agency) whether a consumer is older than 18 years.

1.2 Data ecosystems ensure data storage and access Cloud The cloud becomes the epicenter from which new data (connections), analysis methods, and propositions are developed, tested, and unlocked. In the cloud, data is not only stored; it is also the place where standard and custom analyses, computing power, and unlocking possibilities are offered. This can be from the central cloud or through edge computing, to relieve the central line. For organizations, this offers various advantages to increase flexibility and strength, and to quickly use new technological possibilities (cloud computing services). Cloud parties facilitate the cloud space, but also offer more and more standard analysis services specifically for financial services. At the same time, they offer a strongly secured environment that can meet the increasingly higher requirements from regulation.

OCCASIONAL PAPER Technology Towards 2035 8 Digital twin Creating a digital twin offers potential for monitoring real estate but also for monitoring consumer behavior. A digital twin is nothing more or less than a digital copy of an object or individual. With the digital copy, analyses can be performed to determine what the impact or behavior will be of the digital copy in a changing situation. This offers possibilities to test various fictitious situations on the real estate portfolio or on customers, to then determine how to possibly react. Application programming interfaces (APIs) APIs are software interfaces that make it possible for different applications to interact with each other in a low-threshold manner. With APIs, data can be shared in a low-threshold manner, and full services can be integrated into one's own environment. This forms, among other things, the basis for the development of embedded financial services. For example, Buy Now Pay Later (BNPL) services can be integrated into the payment screen of an online store via an API.

1.3 Data analysis and especially AI are taking off enormously The word 'AI' has existed since the second half of the twentieth century. But in recent years, this technological development has taken off enormously. In the coming years, market parties can, partly due to AI, react more proactively (predictive and preventive) to all kinds of situations. AI has potential in many processes of the chain, with optimization as the main goal. Themes such as explainability and ethics will gain attention. The way of modeling and the fact that the foundation of models is often based on historical data, continue to lead to points of attention. Think here of the chance of extrapolation of historical patterns, being aware of biases, and the fact that models are always a simplification of reality. Issues regarding explainability and ethics remain relevant for every step in the process (input, throughput, output). The AFM has recently, in cooperation with DNB, called attention to this in the report 'The Impact of AI on the Financial Sector and Supervision'. The availability of more and more data points of an individual offers possibilities for segmentation at the micron level. Because there is more and more interaction in the digital world and also behaviors in the physical world are converted into data, it becomes possible to analyze and follow behavior at the individual level. This creates possibilities to personalize and respond to explicit and possibly also implicit needs of individuals. At the same time, it also becomes possible to respond to the weaknesses of an individual and influence them. Generative AI offers various application possibilities for process optimization. The full potential of generative AI - and more specifically large language models (LLMs) – is not yet clear. But it is clear that it can lead to all kinds of new application possibilities. Besides process efficiency in customer service and performing translations, possibilities also arise to automate interaction with consumers.

1.4 Protection of digital security becomes top priority The attention for cybercrime, and the potential risks that follow from it, will increase. Everything connected to the internet is potentially sensitive to hacks. Hackers seem to focus their attention on data environments, due to the value of data. Disrupting systems or processes by adjusting data or