Banks adopt new controls to counter AI-powered scams

Outseer hosted an online panel that examined how major banks are changing consumer scam controls to respond to AI-powered attacks and scam-as-a-service. The session drew on an Outseer survey of consumers across more than 15 markets and set a loss projection for the sector at USD 55.3 billion by 2030. The discussion focused on the threat landscape expected in 2026 and on controls that banks are testing to reduce consumer losses.

The webinar featured Martyn Higson of Outseer and was moderated by Sharon Kimathi. Panelists reviewed survey findings on consumer concerns and expectations, and outlined where technical tools and behaviour-based measures are being combined. The panel discussed how attackers increasingly exploit human behaviour rather than only technical vulnerabilities, and how firms are responding across the lifecycle of an attack.

Speakers organised interventions into five stages of a scam: pre-attack, consumer compromise, transaction, receipt and post-attack. At the pre-attack stage, banks are investing in campaign detection, blocking mechanisms and early warning systems to intercept fraudulent outreach before it reaches customers. During consumer compromise, firms are trialling behavioural prompts, targeted education and added friction in high-risk situations to prevent victims acting on social-engineering requests.

At the transaction stage, banks are using enhanced real-time risk scoring, stricter authentication for high-risk transfers and temporary holds on payments flagged as suspicious. For receipt and confirmation, institutions are testing multi-channel alerts and voice or video confirmation for large or unusual payments. Post-attack measures discussed included faster investigation workflows, clearer remediation paths for victims and reimbursement policies.

Outseer’s cross-market survey was used to compare which controls consumers accept and which controls correlate with reductions in reported loss. The survey data indicated that customers expect banks to both block fraudulent activity and provide clear communications and timely remediation when scams occur. Acceptability of protective measures varied by market and customer segment, according to the panel.

The discussion covered the use of machine learning on both sides: fraudsters use AI to scale more convincing scams, while banks deploy AI-driven detection and analytics to spot anomalies and prioritise responses. Panelists noted the importance of layering algorithmic detection with human review and behaviour-informed interventions.

Panelists recommended continued testing of controls across the five stages, greater application of behavioural science in product design, and ongoing market-level analysis of consumer expectations to inform which measures to keep or modify.