Machine learning
MLAlso called ML
Machine learning is a branch of artificial intelligence in which systems learn patterns from data rather than following fixed rules. Banks and insurers use it to detect fraud, monitor transactions, score credit and screen names. Because it relies on data and statistical methods, it falls under model risk management, data protection and AI rules.
How it works
A machine learning model is trained on historical data, such as past alerts labelled as suspicious or not. It learns which features predict an outcome and scores new cases. In transaction monitoring, firms often run models alongside rules to rank alerts and cut false positives. Supervised models need labelled outcomes. Unsupervised models find anomalies without labels and suit new typologies.
Supervisors treat machine learning as a model. The PRA's SS1/23 covers AI and machine learning models used by banks with internal model approval. In the US, guidance issued in April 2026 replaced SR 11-7 and covers non-generative AI models. Firms must validate models, monitor drift and document data. Credit scoring models also fall within the EU AI Act's high-risk category from 2 December 2027.
The rules
Sets five principles for model identification, governance, development, independent validation and risk mitigants, covering AI and machine learning models.
UK banks, building societies and PRA-designated investment firms with internal model approvalReplaces SR 11-7 with principles-based model risk guidance; generative and agentic AI models are outside its scope.
US banking organisations, most relevant above $30 billion in total assetsBans certain AI practices, sets risk management, data, documentation and human oversight duties for high-risk AI, and transparency duties for chatbots and generated content.
Providers and deployers of AI systems placed on the EU market or used in the EUGives individuals the right not to be subject to solely automated decisions with legal or similarly significant effects, subject to safeguards including human intervention.
Controllers processing personal data of people in the EUCommon questions
How is machine learning used in AML?
Firms use it to score and prioritise transaction monitoring alerts, detect unusual behaviour, link related entities and reduce false positives in screening. Most deploy it alongside rules-based scenarios rather than replacing them.
Is a machine learning model a model under model risk rules?
Generally yes. The PRA's SS1/23 explicitly covers AI and machine learning techniques. US guidance from April 2026 covers non-generative AI models that meet its definition of a model.
What is model drift?
Model drift is a fall in a model's accuracy as real-world data changes from the data it was trained on. Firms monitor performance metrics and retrain or recalibrate models when drift is detected.
