Financial crime and artificial intelligence

Artificial intelligence is rapidly transforming the way financial institutions manage compliance and the risks associated with financial crime. Systems based on machine learning now monitor transactions, identify anomalous patterns and automate regulatory reporting processes across global financial networks. These technologies promise greater operational efficiency and more advanced control capabilities in the fight against illicit financial flows. However, the integration of AI into compliance infrastructures also presents a deeper governance dilemma. When automated monitoring systems operate within particularly punitive regulatory contexts, they can reinforce defensive compliance behaviour and inadvertently contribute to financial exclusion.

Over the past two decades, the global regulatory frameworks against money laundering and terrorist financing have expanded significantly. The international standards developed by the Financial Action Task Force (FATF) have strengthened requirements relating to customer due diligence, transparency regarding beneficial ownership and the monitoring of transactions. These tools are designed to protect the integrity of financial systems and curb illicit financial flows. At the same time, enforcement actions and regulatory sanctions have become more frequent and visible, placing compliance functions at the heart of financial institutions’ governance (Financial Action Task Force, 2025; Bank for International Settlements, 2023). In this context, banks must continuously assess the regulatory risk associated with their customers, the economic sectors and the geographical areas in which they operate.

Artificial intelligence is playing an increasingly central role in this control infrastructure. AI-based monitoring systems analyse large volumes of financial data to identify suspicious patterns and flag transactions for further scrutiny. These technologies dramatically expand financial institutions’ ability to detect potential regulatory breaches. However, they also alter organisational incentives. As the cost of failing to detect suspicious activity can be very high, algorithmic monitoring systems tend to prioritise reducing false negatives – that is, instances where unlawful activity goes undetected. In practice, this means that, in the face of algorithmic uncertainty, risk classification systems may adopt more conservative assessments, classifying even fully legitimate customers or transactions as high-risk.

Within this regulatory and technological context, financial institutions often tend to adopt increasingly prudent compliance strategies. Studies on banking compliance show that organisations frequently interpret regulatory obligations in a defensive manner when regulatory expectations are ambiguous or the potential penalties are very high (de Koker & Symington, 2014). Rather than managing risk through operational controls, institutions may choose to directly reduce their exposure to regulatory risk.

This trend contributes to the spread of ‘de-risking’, whereby financial intermediaries terminate or restrict relationships with clients deemed to be high-risk from a regulatory perspective. Remittance providers, humanitarian organisations and small financial institutions operating in developing countries are among those most affected by these decisions. In many cases, the result is ‘de-banking’, that is, the loss of access to essential banking services by entirely legitimate economic actors (World Bank, 2022; International Monetary Fund, 2023). As de Koker and Casanovas (2024) observe, de-risking demonstrates how regulatory systems designed to combat financial crime can produce unintended consequences when institutions respond to regulatory uncertainty by adopting overly conservative compliance strategies.

The paradox underlying this phenomenon lies in the structure of current regulatory regimes. Anti-money laundering frameworks are based on the principle of risk-based supervision, which encourages institutions to assess and manage risks rather than eliminate them. However, the incentives built into regulatory systems often push in the opposite direction. From the perspective of financial institutions, the asymmetry of consequences is clear: under-compliance can result in legal sanctions, enforcement actions and very significant reputational damage, whilst over-compliance rarely leads to regulatory penalties. Faced with this asymmetry, many organisations adopt defensive strategies that minimise exposure to regulatory risk, even when such strategies reduce financial inclusion or limit economic participation (Organisation for Economic Co-operation and Development, 2021).

Understanding this dynamic requires moving beyond a linear interpretation of regulation and examining the systemic interactions between supervisory authorities, financial institutions and technological infrastructure. From the perspective of complex systems, compliance regimes enhanced by artificial intelligence can generate reinforcing feedback loops that influence the evolution of financial governance.

When regulators step up their supervision in response to the risks of financial crime, financial institutions introduce increasingly sophisticated monitoring technologies and strengthen their internal compliance mechanisms. These technologies generate a growing number of alerts and risk classifications, encouraging increasingly conservative interpretations of the rules. As risk aversion increases, institutions may decide to terminate relationships with clients deemed problematic, thereby exacerbating financial exclusion. Those excluded often seek alternative channels to conduct transactions, such as informal financial networks or unregulated payment systems. When financial activities shift towards these channels, the transparency of the regulated financial system diminishes. Reduced visibility of financial flows fuels further regulatory concerns and may prompt policymakers to tighten monitoring requirements. The result is a self-reinforcing cycle in which regulatory pressure, algorithmic monitoring and defensive compliance feed into one another.

A second reinforcing dynamic emerges from the economic consequences of compliance technologies. AI-based monitoring systems require significant investment in data infrastructure, compliance software and staff specialising in algorithmic risk modelling. These investments increase the overall cost of compliance, particularly for institutions operating across multiple jurisdictions. As costs rise, banks may choose to prioritise clients who are more profitable and have lower exposure to regulatory risk. Smaller clients or organisations operating in complex environments therefore risk losing access to banking services, reinforcing de-risking dynamics and widening financial exclusion (Artingstall et al., 2016).

Regulatory frameworks may, however, also include counterbalancing mechanisms designed to counteract these escalation dynamics. When policymakers recognise that excessive de-risking reduces transparency and limits financial inclusion, supervisory authorities can introduce clearer guidelines or regulatory reforms that encourage institutions to manage risk rather than completely avoid certain sectors. Greater regulatory clarity reduces uncertainty within financial institutions and limits the incentives for overly conservative compliance. By promoting a genuinely risk-based approach, regulators can help restore a more stable equilibrium in the financial system.

The broader lesson for financial governance is that regulatory systems must be assessed not only for the rules they impose, but also for the behavioural feedback loops they generate within technologically mediated financial infrastructures. Artificial intelligence significantly enhances financial institutions’ ability to identify suspicious activity, but it also amplifies risk classification processes and organisational compliance responses. Without careful governance design, these technological capabilities may reinforce defensive strategies that reduce transparency and restrict access to financial services.

Addressing the dilemma of de-risking therefore requires regulatory approaches that combine technological innovation, proportionate supervision and greater regulatory clarity. Aligning regulatory incentives with the objectives of transparency, financial inclusion and responsible risk management will be essential if compliance systems based on artificial intelligence are to strengthen financial governance rather than inadvertently fuelling cycles of over-compliance and exclusion. (photo by Bryan Dijkhuizen on Unsplash)

Artingstall, D., Dove, M., Howell, J., & Levi, M. (2016). Drivers and impacts of de-risking: A study of banks’ de-risking activities. Journal of Financial Crime, 23(3), 527–545. https://doi.org/10.1108/JFC-04-2015-0026

Bank for International Settlements. (2023). Sound management of risks related to money laundering and financing of terrorism. BIS.

de Koker, L., & Casanovas, P. (2024). De-risking, de-banking and denials of bank services: An over-compliance dilemma? In L. de Koker & P. Casanovas (Eds.), Financial crime, law and governance (pp. 45–70). Springer. https://doi.org/10.1007/978-3-031-59547-9_3

de Koker, L., & Symington, J. (2014). Corporate compliance: Reflections on a study of compliance responses by South African banks. Law in Context, 30, 228–256.

Financial Action Task Force. (2025). The FATF recommendations: International standards on combating money laundering and the financing of terrorism and proliferation (Updated ed.). FATF.

International Monetary Fund. (2023). Financial integrity and inclusion: Strengthening the global AML/CFT framework. IMF.

Organisation for Economic Co-operation and Development. (2021). OECD regulatory policy outlook 2021. OECD Publishing.

World Bank. (2022). Financial inclusion and the role of correspondent banking. World Bank.