The fraud threat facing Nigerian financial institutions is changing.
Fraud prevention has traditionally focused heavily on the individual transaction:
- Is this payment unusual?
- Does it breach a rule?
- Should it be blocked?
But the experiences shared by senior fraud prevention and compliance professionals at Regfyl’s second When Fraud Happens executive roundtable suggest that the industry’s biggest challenges increasingly begin before the transaction occurs and extend well beyond the boundaries of any individual institution.
In June 2026, Regfyl convened professionals from across Nigeria’s financial-services ecosystem for the second edition of When Fraud Happens.
The closed-door discussion brought together experiences from commercial banks, microfinance institutions, fintechs, payment companies and other participants confronting fraud every day.
Five themes stood out.
1. Mule Accounts Have Become One of the Industry’s Most Urgent Fraud Problems
Across institution types, mule accounts repeatedly emerged as one of the most pressing fraud challenges. A mule account is an account used to receive, move or withdraw criminal proceeds.
Some are opened deliberately for fraud. Others belong to legitimate customers who knowingly or unknowingly allow criminals to use their accounts.
What makes mule accounts particularly difficult is that they may appear entirely legitimate during onboarding.
A customer may have a valid BVN, a valid NIN, legitimate identification documents, a verified address, and no obvious negative history. Yet once the account becomes active, it may be used as part of a larger network to receive and rapidly move fraudulent proceeds.
One participant captured the problem clearly:
“We have seen accounts that look completely clean at onboarding — verified BVN, utility bills, everything — that are clearly designed to receive and layer funds.”
The issue becomes harder when funds are distributed across multiple accounts. Instead of one obviously suspicious transaction, fraud proceeds may move through a chain:
Origin → Mule A → Mule B → Mule C → Withdrawal
Each individual movement may remain below a rule threshold or appear relatively ordinary when considered alone.
The suspicious pattern only becomes visible when the transactions and relationships are viewed together.
What This Means for Financial Institutions
Mule-account prevention cannot depend exclusively on post-transaction monitoring.
Institutions increasingly need pre-onboarding and early-life intelligence capable of identifying risk before an account becomes deeply embedded in the payment ecosystem. This includes stronger:
identity assurance; device intelligence; behavioural analysis; related-party analysis; account-linking intelligence; early-life account monitoring; and cross-institution fraud intelligence.The strategic shift is from: “Is this identity valid?” to: “Is this identity, device and account likely to behave legitimately?”
2. APP Fraud Is Creating New Questions About Responsibility
Authorised Push Payment fraud occurs when a genuine customer is deceived or manipulated into authorising a payment to a fraudster.
This could arise through an investment scam, romance scam, business email compromise, impersonation of a bank or government agency, or another form of social engineering.
APP fraud is particularly difficult because the transaction itself may pass every traditional authentication check.
The real customer logs in. The real customer initiates the payment. The real customer enters the authentication credentials. The real customer presses send.
Yet the customer is acting because someone has manipulated them.
That creates a very different fraud problem from account takeover or stolen credentials.
Participants broadly supported stronger protection for customers, but the discussion also revealed significant uncertainty around how reimbursement frameworks should work in practice.
If a customer was deceived into making a payment, when should the sending institution bear responsibility? What responsibility should sit with the institution receiving the funds? What evidence should be required before reimbursement? How should institutions distinguish genuine victims from fraudulent or opportunistic claims?
There was no universal consensus.
But there was broad agreement that any future framework needs to balance customer protection with clear evidential standards and appropriate incentives for all participants in the payment chain.
Fraud Prevention Must Happen Before Authorisation
APP fraud also exposes the limits of authentication as a fraud control.
Authentication answers the question: “Is this the legitimate customer making the transaction?”
It does not necessarily answer: “Is this a transaction the customer would ordinarily make if they were not being manipulated?”
That distinction makes behavioural intelligence increasingly important.
A new beneficiary, an unusual payment amount, a sudden device change, atypical transaction timing or a payment completely inconsistent with previous behaviour may provide useful context even where the customer has successfully authenticated.
APP fraud therefore pushes fraud prevention beyond verifying identity towards understanding intent, behaviour and context.
3. Strong KYC Remains One of the Most Effective Fraud Controls
One of the more striking themes from the roundtable was the continued importance of Know Your Customer (KYC) controls.
When participants discussed which controls had the greatest impact on fraud prevention, strong onboarding and identity assurance repeatedly came to the forefront.
That is significant because much of the industry conversation around fraud prevention focuses on what happens after an account is already active: transaction monitoring, behavioural analytics, fraud scoring and real-time intervention.
But participants repeatedly described cases where the real control failure happened much earlier.
The lesson is not that transaction monitoring is less important.
It is that a transaction-monitoring system is operating at a disadvantage if the institution has already onboarded the wrong person, relied on compromised identity information or failed to identify meaningful connections between apparently unrelated accounts.
KYC Must Move Beyond Document Collection
Effective KYC for fraud prevention therefore needs to go beyond confirming that a customer possesses valid documents.
Institutions increasingly need greater confidence that the person controlling the account is genuinely the person represented by the identity, that the identity is not synthetic or compromised, and that apparently unrelated customers are not connected through common devices, addresses, beneficiaries or other attributes.
The customer’s early behaviour after onboarding also matters.
A new account that immediately begins receiving funds from multiple unrelated individuals and rapidly dispersing those funds may deserve attention even where every onboarding document was technically valid.
This shifts the objective from basic KYC towards identity assurance.
The institution is not simply asking whether documentation exists. It is asking whether the customer behind the account is genuine, understood and behaving consistently with what was represented at onboarding.
4. Fraud Intelligence Sharing Is Becoming Essential
Fraudsters do not operate within the boundaries of individual institutions.
A scam may start with a victim at one bank, move into an account at another bank, pass through a fintech wallet or payment processor, split across several additional accounts and eventually leave the formal financial system altogether.
Every institution may see only one part of the journey. The fraudster sees and orchestrates the whole thing. That creates a fundamental information disadvantage.
Participants broadly agreed that greater fraud-intelligence sharing is becoming essential, even though significant legal, privacy, operational and governance questions remain.
The goal is not uncontrolled sharing of customer information. It is the development of lawful and structured mechanisms through which institutions can share useful indicators, emerging typologies and information about patterns that are already moving across the ecosystem.
That could include intelligence about mule accounts, recurring beneficiaries, new fraud typologies, account patterns or other risk indicators capable of helping institutions recognise threats earlier.
The value is speed.
If one institution identifies a new fraud pattern today but the rest of the ecosystem learns about it months later, fraudsters have a large window in which to repeat the same method elsewhere.
A stronger intelligence-sharing environment can shorten the time between emergence, detection and industry-wide response.
5. Some of the Biggest Fraud Problems Are Ecosystem Problems
Perhaps the most important conclusion from the roundtable was that several major fraud challenges are not problems any single financial institution can solve on its own.
They are structural.
Three issues illustrated this particularly clearly.
Data Integrity
Financial institutions depend on national identity and verification infrastructure to support customer onboarding.
Where underlying identity information is incomplete, inconsistent, outdated or compromised, those weaknesses flow directly into institutional KYC and fraud-prevention processes.
An individual bank or fintech can strengthen its own onboarding controls, but it cannot independently repair the underlying national infrastructure.
That means some fraud-prevention improvements require engagement at an ecosystem level rather than simply further investment inside one institution.
Staff-Enabled Fraud
Participants also discussed the role of insider facilitation.
Staff-enabled fraud can involve fraudulent account creation, deliberate circumvention of controls, inappropriate access to customer information, data leakage, collusion or the facilitation of suspicious transactions.
The issue cannot be solved purely through better recruitment.
It requires stronger access controls, segregation of duties, monitoring of privileged activity, accountability and effective internal governance.
It also demonstrates why fraud risk should not be treated solely as a customer problem.
The threat can originate from outside the institution, from customers themselves or from people inside the organisation.
Post-Fraud Intervention
The period immediately after fraud is detected can have a major impact on whether funds are ultimately recovered.
But participants highlighted persistent operational challenges around account restrictions, fund recall, inter-institution communication, evidence gathering and engagement with law enforcement.
A transaction may be detected quickly, but if the response process is slow, the proceeds can move across several institutions before meaningful action occurs.
That means the strength of a fraud framework should be assessed not only by how quickly it identifies a problem, but by how quickly the institution and the wider ecosystem can act on that information.
The gap between detection and intervention may represent one of the most important opportunities for reducing fraud losses.
Three Questions the Industry Is Still Trying to Answer
Not every issue discussed during When Fraud Happens II produced a clear answer.
That is part of what made the discussion valuable.
The roundtable highlighted several areas where technology, regulation and operational practice are still evolving.
How Early Should Institutions Intervene?
Participants debated whether financial institutions should intervene before a fraudulent transaction has fully materialised.
Behavioural intelligence may suggest that an account or transaction presents significant risk even where there is not yet definitive evidence of fraud.
Earlier intervention could prevent losses.
But it also raises difficult questions about customer rights, false positives, contractual authority, consistency and the evidence required to restrict legitimate activity.
The general direction of the conversation favoured earlier intervention where risk signals are strong, but with documented and proportionate escalation procedures.
The challenge is to become more proactive without turning every behavioural anomaly into an unjustified restriction.
Can Machine Learning Outperform Traditional Rules?
Participants also discussed growing use of machine-learning-assisted anomaly detection.
Some institutions reported positive experiences, particularly around reducing false positives and identifying behavioural patterns that conventional rules could miss.
Others highlighted significant concerns around explainability, model drift and the ability to defend decisions to regulators.
No clear winner emerged between rules and machine learning.
The more likely direction is a hybrid model.
Rules remain useful for known, explainable and clearly defined risk scenarios. Adaptive models can add value in identifying patterns and anomalies that are difficult to capture using static thresholds alone.
The important question may therefore not be whether institutions should choose rules or machine learning.
It may be how the two can work together while maintaining appropriate governance and explainability.
Who Should Bear the Cost of APP Fraud?
Liability for APP fraud remained one of the most contested areas of discussion.
Sending institutions, receiving institutions and customers all play different roles in the payment journey, and participants naturally had different views on where responsibility should sit.
There was, however, some support for models that recognise shared responsibility rather than automatically placing the full burden on one party.
Any future Nigerian framework will need to consider not only customer protection but also how liability rules affect behaviour.
Poorly designed incentives could encourage weak controls on one side of the transaction or fraudulent claims on the other.
The evidential framework will therefore matter just as much as the reimbursement principle itself.
The Future of Fraud Prevention Will Be Won Before the Transaction Happens
Taken together, the findings from When Fraud Happens II point towards a broader shift in fraud prevention.
The traditional model often begins with the transaction:
Transaction → Alert → Investigation → Response
The emerging model begins much earlier:
Identity → Intelligence → Behaviour → Detection → Intervention → Response → Learning
Identity assurance matters because fraud prevention becomes harder once a mule or compromised identity is already inside the institution.
Intelligence matters because fraudsters operate across institutions and exploit information gaps.
Behaviour matters because authentication alone cannot determine whether a legitimate customer is acting under manipulation.
Rapid response matters because detecting fraud without the ability to intervene quickly may still result in lost funds.
And governance matters because effective fraud prevention requires clear accountability, senior-management visibility and documented escalation processes.
Technology remains central to all of this.
But the discussion also reinforced a more fundamental point:
“Many of the industry's biggest fraud challenges are no longer technology problems. They are challenges of trust, coordination, accountability and speed of response.”
The institutions best positioned for the future will therefore not necessarily be those with the largest number of fraud rules.
They will be the institutions with stronger intelligence, better onboarding controls, faster response capabilities and deeper collaboration networks.
Download the Full When Fraud Happens II Insights
The full When Fraud Happens II report captures the detailed findings from Regfyl’s June 2026 executive roundtable and explores the issues behind the discussion in greater depth.
It covers mule-account layering, APP fraud, KYC and identity assurance, intelligence sharing, data-integrity weaknesses, staff-enabled fraud, post-fraud intervention, behavioural intelligence, machine learning and the questions the industry has yet to resolve.
Download the When Fraud Happens II Insights
For insights from the inaugural roundtable on fraud response, governance, product design, law-enforcement engagement and the AFIU Fraud Framework:
Every mule account we have ever seen passed our monitoring systems because it looked like a legitimate account from a legitimate customer. The problem started at onboarding. Participant - WFH II
The fraudsters know exactly which bank to hit next. We are the only ones operating without a complete picture. Participant - WFH II