Banks Adopt Layered Controls to Fight AI-Driven Scams

An online webinar hosted by Finextra and Outseer brought together industry practitioners and researchers to review consumer scam controls and new survey data. The event focused on consumer scam concerns for 2026, the controls banks use at each stage of an attack, and survey findings from more than 15 markets. Martyn Higson, principal product manager at Outseer, spoke on the panel and Sharon Kimathi of Finextra moderated.

Panelists reported a rise in the scale and sophistication of scams and cited industry estimates that losses to financial institutions could reach $55.3 billion by 2030. Panelists noted many recent campaigns target human behaviour rather than technical vulnerabilities, shifting attention to consumer-facing warnings and interventions.

Speakers organised controls into five stages of a scam: pre-attack, consumer compromise, transaction, receipt and post-attack. At the pre-attack stage they described threat intelligence and early detection to identify criminal infrastructure and risky channels. During consumer compromise the discussion covered real-time warnings, user education and behavioural nudges to interrupt social-engineering flows. For the transaction stage panelists examined transaction monitoring, risk scoring and decisioning that can pause or challenge suspicious transfers. Receipt-stage controls discussed included payee confirmation and reconciliation checks. Post-attack responses focused on rapid investigation, customer remediation and recovery processes.

Panelists discussed combining automated detection and blocking with behavioural science to reduce the success of social-engineering attacks. The Outseer survey was used to report consumer attitudes toward bank intervention and recovery support, indicating which measures customers are more likely to accept.

Speakers addressed operational challenges for banks, including balancing security friction with user experience, coordinating across fraud, payments and customer service teams, and scaling human oversight where automated systems flag ambiguous cases. Panelists recommended sharing anonymised threat intelligence across institutions and updating detection models as attackers adopt more advanced tools.

Organisers outlined areas for further work including faster detection, refining behavioural interventions, and aligning reimbursement and customer-support practices with consumer expectations captured in the multi-market survey.