As account-to-account payments move deeper into mainstream digital banking, fraud prevention is becoming a race against increasingly coordinated scams. Visa is expanding A2A Protect with real-time risk intelligence and a new unified fraud score, while developing Visa Graph IQ, an agentic investigation capability designed to help banks identify fraud networks, money mules and emerging threats before losses spread across the payment ecosystem.
Account-to-account payments are becoming a major part of the global digital-payments infrastructure. That growth is creating a parallel challenge for banks: detecting sophisticated fraud quickly enough to stop a transaction before the money disappears.
Visa is responding with an expanded version of A2A Protect, its fraud-prevention service for account-to-account payments. The enhanced platform adds real-time risk insights and a unified fraud score designed to give financial institutions a faster signal about potentially fraudulent transactions.
The update also represents Visa’s first in-market integration of Featurespace technology, following Visa’s acquisition of the fraud and financial-crime prevention company in 2024.
Visa is separately developing Visa Graph IQ, a graph-powered, agentic investigation capability intended to help fraud and risk teams uncover connections between suspicious transactions, identify money-mule activity, detect coordinated fraud networks and accelerate investigations.
Together, the initiatives point to a broader change in payment security: fraud detection is moving from isolated transaction scoring toward ecosystem-level intelligence that can identify relationships between accounts, transactions and actors.
A2A growth is expanding the fraud surface
The stakes are particularly high in Asia Pacific, where account-to-account payments have become deeply embedded in digital commerce and banking.
Visa says global A2A transactions are projected to exceed 5.8 trillion by 2028, representing a 160% increase from 2024. Asia Pacific is expected to account for more than half of global consumer A2A transactions by then.
The growth is occurring alongside a significant scam problem. Visa cites estimates that Asia Pacific represents approximately 67% of the world’s $1.03 trillion in annual scam losses, while Asia recorded $688.42 billion in scam-related losses during 2024.
Unlike many traditional card transactions, A2A payments can move funds directly between bank accounts. Once an authorized transfer is completed, recovering the money can be difficult.
That makes prevention particularly important.
The challenge is also changing the role of fraud teams. A suspicious transaction may not look obviously fraudulent in isolation. It can become suspicious when considered alongside other accounts, transaction patterns or known scam activity elsewhere in the financial ecosystem.
Moving fraud detection upstream
A2A Protect is designed to address that problem by giving participating banks access to Visa’s broader network intelligence.
The service uses AI and transfer-learning techniques to provide risk signals without requiring a bank to wait months for its own transaction history to generate enough data for a fraud model. Banks can also opt into additional network-level intelligence designed to identify threats emerging across the wider ecosystem.
That distinction could be important for smaller financial institutions.
A bank operating alone may have limited visibility into a fraud campaign because the activity is distributed across multiple institutions. A network-level system can potentially recognize patterns that individual banks cannot see.
Visa says A2A Protect has been shown to increase fraud detection by up to 75% during the first six months of deployment. That figure is a company-reported performance claim rather than an independent industry benchmark, so financial institutions should evaluate the underlying methodology and performance against their own fraud data before making adoption decisions.
Featurespace adds another layer of AI-based risk scoring
The new unified fraud score incorporates Featurespace technology into Visa’s A2A fraud offering.
Featurespace has specialized in adaptive behavioral analytics for detecting fraud and financial crime. Bringing that technology into A2A Protect gives Visa another component for evaluating transaction risk in real time.
For banks, a unified score can simplify a workflow that otherwise requires fraud analysts to interpret multiple risk signals from different systems.
Visa says each A2A Protect alert also includes a plain-language explanation of why a transaction was flagged.
That is an important operational detail. Fraud detection systems do not exist in isolation; analysts need to understand why a transaction has been classified as risky before deciding whether to intervene.
Explainability can also help financial institutions tune fraud controls. If a system generates too many false positives, analysts can identify which signals are producing unnecessary alerts and adjust their processes.
Graph technology takes fraud investigations beyond individual transactions
Visa Graph IQ takes a different approach.
Rather than focusing primarily on whether one transaction looks suspicious, graph-based analysis can examine relationships between entities. Accounts, transactions, merchants, devices and other signals can be represented as connected nodes, allowing investigators to look for patterns across a network.
This is particularly relevant to money-mule networks, where criminals use multiple accounts to receive, move or layer illicit funds.
Visa describes Graph IQ as an agentic investigation capability. In this context, agentic AI can help automate parts of the investigative process—such as navigating relationships, surfacing relevant connections and accelerating analysis—while fraud professionals remain responsible for decisions and actions.
The approach resembles a broader trend across financial services. Banks and fintech companies are increasingly combining machine learning, graph analytics and AI agents to move from reactive fraud alerts toward proactive financial-crime investigations.
The competitive landscape is changing
Visa is not alone in pursuing network-level fraud intelligence.
Mastercard, major banks, payment processors and specialist fraud-technology vendors are investing in behavioral analytics, transaction monitoring, graph databases and AI-powered financial-crime detection.
The differentiator for Visa is its position within a global payments ecosystem. Its network can provide signals that may be difficult for an individual bank to reproduce.
But network intelligence also raises questions around data sharing, privacy, governance and interoperability. Banks must determine what information they are comfortable contributing to shared intelligence systems and how those signals are used.
For enterprise fraud teams, another consideration is integration. A2A Protect connects to existing financial-institution systems through a single API, according to Visa. That could reduce the technical burden associated with deploying another fraud layer, particularly for institutions operating complex legacy payment environments.
What banks should consider
The most important shift in Visa’s strategy is conceptual.
Fraud prevention is increasingly becoming a network problem rather than a transaction problem.
A bank can score a transaction based on its own customer history. A network can potentially recognize that the same destination account, device or transaction pattern is connected to activity appearing across multiple institutions.
For banks expanding real-time payments and A2A capabilities, that distinction could become increasingly important.
The technology will not eliminate scams. Fraudsters can adapt their behavior, exploit legitimate credentials and manipulate customers into authorizing transactions themselves. Detection models must therefore evolve alongside attack techniques.
For enterprise payment teams, the emerging benchmark is likely to be a combination of speed, accuracy and context: detecting risk before authorization, minimizing false positives for legitimate customers and giving investigators enough intelligence to understand coordinated activity.
Visa’s expanded A2A Protect and developing Graph IQ show where that market is heading. The future of payment security may depend less on asking whether one transaction looks fraudulent and more on understanding where that transaction fits within a much larger network of financial behavior.
Market Landscape
The fraud-prevention market is shifting as real-time payments, A2A transfers and digital banking expand.
Traditional rules-based systems remain important, but banks increasingly need behavioral analytics, machine learning, graph analysis and network intelligence to identify coordinated fraud.
Visa’s strategy combines several layers:
- Real-time transaction risk scoring
- Network-level fraud intelligence
- Featurespace behavioral analytics
- Graph-based investigation
- Agentic AI for investigative workflows
- API-based integration with existing banking systems
The competitive landscape includes Mastercard, Featurespace, Feedzai, NICE Actimize and other fraud and financial-crime technology providers. The differentiator for payment networks is access to broader transaction intelligence, while specialist vendors often compete on analytics, orchestration and flexibility across payment rails.
For banks, the strategic question is increasingly whether their fraud stack can detect risk before authorization, rather than simply investigate losses after money has moved.
Top Insights
- Visa’s enhanced A2A Protect adds real-time fraud scoring, helping banks identify suspicious account-to-account transfers before funds leave customer accounts.
- Featurespace technology strengthens Visa’s fraud platform, bringing behavioral analytics into a unified risk score designed to improve detection while reducing unnecessary alerts.
- Visa Graph IQ applies graph technology and agentic AI to help investigators identify fraud networks, money mules and coordinated financial-crime activity.
- Asia Pacific’s rapid A2A growth increases urgency, with the region expected to represent more than half of global consumer A2A transactions by 2028.
- Network-level intelligence could reshape bank fraud operations, giving institutions visibility into cross-bank patterns that may remain invisible within individual transaction datasets.
Get in touch with our fintech expert
