FINMA Accelerates AI-Powered Supervision
Marlene Amstad, Chair of the Swiss Financial Market Supervisory Authority (FINMA) and Chair of the International Organization of Securities Commissions (IOSCO), recently announced that global regulators are rapidly adopting artificial intelligence (AI) and regulatory technology (RegTech) tools to address the growing complexity of crypto market supervision. As Switzerland's primary crypto regulator, FINMA has taken the lead by deploying two AI-driven initiatives specifically designed for crypto oversight.
Real-Time Crypto Monitoring Dashboard: Detecting Concentration Risks
FINMA has developed a real-time crypto monitoring dashboard that integrates quarterly reported crypto asset holdings from institutions with daily market price data. This system dynamically tracks the concentration of crypto exposure within individual institutions. When an institution's crypto position becomes disproportionately large relative to its overall portfolio, or when a token is excessively concentrated on a single blockchain, the dashboard automatically triggers alerts. The former indicates heightened credit risk from concentrated holdings; the latter signals operational risk—for instance, if that blockchain experiences an outage or attack. This real-time capability shifts supervision from a reactive, post-hoc model to proactive risk detection.
Generative AI Tool: Automated Document Screening with Dual-AI Validation
Amstad further disclosed that FINMA is developing a generative AI tool to scan regulatory documents submitted prior to on-site inspections and automatically flag anomalies. The workflow involves two AI layers: the first AI reads through extensive compliance documents and identifies potential issues or unusual statements; the flagged points are then reviewed by a second AI model to cross-validate the findings. Only after this dual-AI verification are the results passed to human regulators for final decision-making. This 'AI prescreening plus AI review' architecture aims to reduce manual review costs while improving the accuracy of anomaly detection. Given Amstad's dual role as IOSCO Chair, this development may signal a push toward standardized application of similar AI tools across international regulatory bodies.