Sanctions screening programmes are generating more alerts than ever, but more alerts do not necessarily mean better compliance. This session looks at how firms can reduce false positives, improve screening decisions and use AI without losing control.

In this session, Spencer Vuksic of Castellum.AI explored why false positives arise, how screening thresholds, data quality and policy gaps contribute to alert fatigue, and what a defensible review and closure process should look like.

The discussion also looked at the growing use of AI agents in AML and sanctions screening, including how they can support alert investigation, documentation and escalation while maintaining explainability, auditability and human oversight.

In this session we covered:

• How sanctions and AML screening systems generate alerts
• The main causes of false positives and screening fatigue
• How better customer data and identifiers can reduce unnecessary alerts
• When to close an alert and when further escalation is required
• What defensible, audit-ready screening documentation looks like
• How AI agents can support AML/KYC alert investigation and adjudication
• How to assess AI tools for explainability, accuracy and operational fit
• The importance of governance and human oversight when introducing AI into compliance workflows

The key message: effective screening is not about clearing more alerts — it is about making better, consistent and well-documented decisions while focusing compliance resources on genuine risk.