Raising Standards in Compliance: Application of artificial intelligence to online data to identify anomalous behaviours

A collaboration between City, University of London, Kindred, and BetBuddy produced a white paper on using AI to identify anomalous behaviours.

This whitepaper presents insights from industry stakeholders on the application of artificial intelligence (AI) to enhance anti-money laundering (AML) compliance in online gambling. It explores regulatory challenges, criminal typologies, and the potential of machine learning to detect anomalous behaviors.

Key Findings:

• Criminal Adaptability: Traditional rules-based systems are insufficient as criminals evolve tactics to bypass static thresholds and exploit regulatory gaps.
• Compliance Culture: Effective AML compliance requires leadership commitment across departments, especially from boards and senior management.
• Operational Challenges: Manual processes for submitting Suspicious Activity Reports (SARs) and limited feedback from crime agencies hinder proactive monitoring.
• Data Sharing Limitations: Legal constraints, particularly GDPR, restrict operators from sharing customer data, though centralized databases are proposed.
• Technology Investment: AI and anomaly detection models (e.g., LSTM, HTM) are essential for real-time monitoring and identifying suspicious behavior below regulatory thresholds.
• SOF/SOW Verification: Enhanced source of funds and wealth checks are critical but must be balanced with customer experience and privacy concerns.
• Industry Collaboration: Joint efforts through groups like GAMLG and partnerships with academia (e.g., City University of London, Kindred Group) are vital for innovation.
• Next Research Phase: Focuses on unsupervised learning models for anomaly detection and improving interpretability of deep neural networks.

This research underscores the need for smarter, adaptive compliance systems and collaborative innovation to stay ahead of financial crime in gambling.

For full insights and technical recommendations, please refer to the complete document.