Banks that opt into Visa’s upgraded A2A Protect will gain earlier detection of account-to-account transfers flagged as risky, Visa says. The company rolled out a unified fraud score that blends Visa’s network signals with Featurespace technology, marking its first in-market integration since acquiring the firm. Visa claims the new score detects over 50% more fraud while trimming unnecessary alerts by over 40%.
Visa framed the update against rapid growth in direct bank transfers, citing a projection that A2A transactions will surpass 5.8 trillion by 2028, a 160% increase from 20241. To handle that scale, the firm says the enhancement applies advanced AI and transfer learning to produce near-instant risk insights across jurisdictions, removing the need for months of in-house model training or waiting on consortium datasets to mature.
Participating financial institutions may add network-level signals to their decisioning, which Visa says helps surface coordinated, emerging threats that individual banks could miss. The product is intended to slot into existing systems through a single API, a design choice aimed at cutting implementation time and operational complexity. Alerts arrive with plain-language explanations of why a transaction was flagged, intended to speed triage by fraud teams while reducing friction for legitimate customers.
“Fraudsters move fast across payment types, and financial institutions need risk insights just as quickly, without slowing down legitimate payments,” said Walter Lironi, Senior Vice President, Head of Value-Added Services, Central and Eastern Europe, Middle East, and Africa (CEMEA), Visa.
Visa positions A2A Protect as a ready layer of defence for institutions that sign on, delivered without the months-long modelling timelines traditionally required. The company directs readers to Visa.com/security for additional detail. What determines the upgrade’s reach and accuracy, Visa stresses, is adoption: banks must opt in to both share and receive the network signals that underpin the unified score.
