Fraud is often described as a growing problem for financial institutions, but in reality, that framing no longer reflects what is happening.
What we are seeing today is not simply more fraud taking place, but a shift in how it operates. Fraud has become faster, more organised and significantly more scalable. It has moved beyond opportunistic activity into something that increasingly resembles a structured industry.
The scale of the issue is overwhelming as global scam losses have reached an estimated $442 billion in just one year, which is only the tip of the iceberg, with the majority of adults experiencing at least one attempt and nearly a quarter losing money. These figures point to a model of financial crime that has changed at its core. Behind that staggering figure for global scam losses lies an uncomfortable truth: banks green-lit every single one of those transactions, unwittingly funnelling that money directly to criminals.
For banks, the implication is clear: if fraud has industrialised, existing approaches to prevention are no longer sufficient.
From Volume to Scale
For years, fraud was understood primarily in terms of volume. More attacks required more controls, and institutions responded by refining detection rules and strengthening authentication and transaction monitoring. Yet today, the defining characteristic is scale.
Fraudsters now operate with a level of coordination and efficiency that mirrors legitimate businesses. They test and refine their methods, reuse successful tactics and share intelligence across networks. Campaigns are designed, measured and optimised for performance.
Artificial intelligence (AI) is accelerating this shift, but it is important to know that it is not the root cause; it simply enhances what already exists. It allows faster targeting, more convincing communication and greater personalisation, which is a devastating combination for victims. Messages that once contained obvious warning signs are now increasingly indistinguishable from legitimate interactions, while campaigns that once took hours can now be executed in minutes.
The Limits of Traditional Controls
This development in fraud tactics exposes a structural gap in how fraud is still being addressed. Most fraud prevention frameworks remain focused on authentication and are designed to detect unauthorised access, identify compromised credentials and block suspicious transactions. That approach still matters, but it is no longer enough.
Modern scams are built around manipulation, and they follow a consistent pattern. Emotion is triggered first, often through urgency, fear, or trust. The victim is then guided through a sequence of steps designed to bypass safeguards. Finally, the payment is authorised by the customer themselves.
From a system perspective, the transaction appears legitimate; authorisation succeeds; the payment is confirmed, and there is no obvious breach. Yet the fraud has already taken place.
This is the defining challenge of authorised push payment (APP) scams. The system is working as intended, but it is being used against itself. Without an understanding of context and intent, traditional controls are left reacting too late. Effective prevention must now operate on both sides of the payment: outbound monitoring protects the payer, but detecting and disrupting mule accounts on the receiving side has long been the structural gap.
Speed Defines Success and Safety
Timing has become one of the most important factors in modern fraud, but it operates across two distinct phases. The manipulation phase is increasingly prolonged, with grooming-based scams such as romance baiting schemes unfolding over weeks or even months as trust is carefully built. By contrast, once a payment is initiated, execution is rapid. Funds are transferred and then quickly moved through mule networks, converted into other assets or withdrawn, often within minutes.
At the same time, payment systems are becoming faster and more interconnected. Instant payments are expanding globally, reducing the window for detection and intervention. However, fraudsters are adapting in step. They create urgency, compress decision-making, and push victims towards immediate action. In doing so, they exploit the very speed that modern payment infrastructure is designed to deliver. For financial institutions, this makes real-time detection and response essential, and delayed intervention is no longer an option.
The Human Layer Remains the Primary Target
Despite advances in technology, fraud will continue to begin with people. Social engineering remains at the core of most attacks, and the uncomfortable reality is that the human has become the weakest link in the security chain. Increasingly, the point of vulnerability is not within the bank's infrastructure, but at the customer level, where individuals are being systematically manipulated into authorising transactions themselves.
AI has removed many of the traditional warning signs: poor grammar, inconsistent tone and obvious errors are all becoming less common. In their place are highly personalised and context-aware interactions that feel credible and urgent.
Fraudsters are not just trying to deceive; they are looking to systematically influence behaviour. As a result, the distinction between legitimate and fraudulent interaction is becoming increasingly difficult to identify, both for consumers and for professionals within financial institutions.
Regulation is Changing Incentives
Regulators are beginning to respond to this shift. In the UK, mandatory reimbursement for APP fraud, effective from 7 October 2024, has placed responsibility on both sending and receiving institutions for scams executed via Faster Payments and CHAPS, fundamentally altering incentives. Across Europe, the forthcoming PSD3 and the Payment Services Regulation (PSR) will extend similar expectations, requiring payment service providers to reimburse victims of APP fraud in cases such as impersonation scams, where prevention standards have not been met.
These developments are important. They reinforce the need for stronger solutions and greater accountability. However, regulation does not prevent fraud; it determines how the consequences are shared. The underlying challenge remains unchanged. Institutions must detect and stop fraudulent activity before funds leave the system.
From Isolation to Collaboration
One of the clearest lessons from the current landscape is that fraud cannot be addressed in isolation. Fraudsters operate as networks, sharing infrastructure, data and tactics. When one institution blocks an approach, it is quickly adapted and deployed elsewhere. Banks, by contrast, have traditionally operated within their own data environments, creating a serious disadvantage.
Many of the most valuable fraud signals exist beyond a single institution. Suspicious beneficiary accounts, mule networks, behavioural anomalies, and emerging scam patterns are all part of a broader ecosystem. When this intelligence is shared and applied in real time, both detection and investigation improve significantly. When it is not, fraud simply shifts to the next institution.
This is why collaborative approaches to fraud detection are becoming increasingly important, enabling institutions to benefit from shared insight rather than isolated data sets. In practice, this is already taking shape through network-level intelligence services such as EBA CLEARING's Fraud Pattern and Anomaly Detection (FPAD) for SEPA, as well as community frameworks like Community Scoring & Intelligence, which allow participants to identify and respond to threats collectively.
A New Baseline for Fraud Prevention
Addressing modern fraud requires a shift in approach. Static rules and post-event analysis are no longer sufficient in an environment where fraud evolves in real time. Instead, institutions need to focus on behaviour, context and real-time decision-making.
This means moving towards systems that can assess risk dynamically, identify anomalies as they emerge, and intervene before transactions are completed. It also means extending visibility beyond internal systems to the wider payment ecosystem, across all payment rails. This requires a shift from isolated systems to interoperable solutions that can ingest external intelligence and apply risk scoring across both internal and network-level signals in real time.
Fraud has become an organised and scalable industry. To respond effectively, financial institutions must match that level of coordination, intelligence and speed. The challenge is no longer simply to detect fraud after it occurs, but to prevent it before it succeeds.