Governance

Why another payment approval may not stop supplier fraud

Several approvers can rely on the same fraudulent email. Before funding another payment check, establish what new evidence it provides and test whether it can be bypassed.

A supplier emails your finance team asking for its next invoice to be paid into a different bank account. Two employees approve the change. If the email came from a fraudster, both employees may have approved sending the money to the wrong person.

Requiring a third approval would not necessarily prevent that loss. If all three people rely on the same fraudulent email, none has established that the real supplier requested the change.

For the CFO considering stronger payment controls, the question is what the proposed check will establish that the existing approvals do not. A documented fraud case shows why that question reaches beyond the payment itself.

The fraud began before the bank details changed

In a case published by the Audit Office of New South Wales in October 2025, an impersonator asked a government agency to change a supplier’s email address. The agency’s procedure required checking the request through the email address already on file. Staff did not follow that procedure. They accepted the change after a phone call with the impersonator.

The fraudster then used the new address to obtain copies of genuine supplier invoices.

That mattered because the agency’s checks for changing a supplier’s bank details required three current invoices and an email address matching its supplier records. The fraudster could now satisfy both conditions. The agency approved a payment to the fraudulent account; its bank subsequently reversed the payment.

The later checks were not examining independent evidence. The attacker had already changed the contact record they relied on. Genuine invoices made the request look credible without establishing who had authorised the new bank account.

The agency responded by requiring staff to contact the supplier’s accounts team using a number obtained independently from the supplier’s website, and by introducing a review of whether the verification steps had been completed.

This case demonstrates a particular failure, not how common it is or the financial return from a proposed remedy. The bank’s reversal is also an outcome of this incident, not something another firm can assume will happen. The lesson for a control proposal is to examine how a supposedly trusted contact became trusted, and what prevents an attacker from changing it.

Give the second approver something different to check

Payment approval still has a purpose. It can establish that an expense is authorised, catch an entry error or confirm that required checks have been performed. The problem arises when those approvals are also treated as proof that a supplier’s instruction is genuine.

Consider two ways of strengthening the process. In the first, another employee reviews the same email and invoices. They may spot an inconsistency, but if the documents are convincing, they can reach the same mistaken conclusion.

In the second, an employee confirms the specific bank-account change with an authorised supplier contact reached through a route established independently of the request. Another reviewer then checks the evidence of that confirmation and that the confirmed account is the account prepared for payment. That review has something new to examine.

The contact details need protection too. Calling a number supplied in the disputed email does not provide independent confirmation. Nor can staff simply assume that a number in their own records is safe if someone has recently changed it. Existing records and supplier websites can themselves be compromised; neither is an automatic guarantee of authenticity.

Before buying another layer of approvals, establish what the current approvers actually do. If they already obtain and review independent confirmation, the proposed step may duplicate it. If the procedure already requires verification but staff bypass it, another written rule will not repair the failure to carry it out. The NSW case began with just such a failure.

Will the check survive a payment deadline?

Imagine a bank-detail change arriving at 3.30 pm, with the payment cut-off at 4 pm. The established supplier contact cannot be reached. This is an illustrative situation, not another reported incident.

If the new check is intended to prevent payment to an unverified account, the payment to that account stays on hold. A finance manager can seek another trustworthy route to an authorised supplier contact and address the consequences of delay. Approving an urgent exception does not, by itself, establish that the account is genuine. Using the old account is not automatically safe either: it may genuinely have closed.

This is where a proposal acquires an operating cost. Someone must maintain supplier contacts, perform the checks, cover absences and handle failed attempts. Delays can mean lost discounts, late fees or interrupted supply where those consequences apply. A process that routinely ends in an unverified manual payment has not resolved the original problem.

New Zealand’s National Cyber Security Centre recommends testing business-process verification with a small group to understand how often checks are needed and how much time they add. It also warns that processes poorly fitted to normal operations may be avoided.

For the CFO, that suggests a practical requirement before approving the budget: ask the proposal owner to observe representative requests, including unsuccessful contact attempts. Record staff time separately from elapsed waiting, identify who would provide cover, and establish what payment delays would actually cost. Include contact maintenance and changes to payment permissions in the proposal, not just the time spent making a call.

Those observations can inform the operating budget. They do not establish an annual fraud probability or justify an invented percentage reduction in losses.

Ask to see a payment that should be stopped

Approving the design and budget is different from accepting the control as effective.

Before relying on the implemented change, ask the operations owner to demonstrate a legitimate bank-account change and an attempted fraudulent one through the configured payment process. Include a recently altered supplier contact, an unavailable contact near cut-off and an attempt to release the payment manually.

Watch which contact staff use, what they confirm and which bank account is prepared for release. Check that a legitimate change can proceed once verification and approval are complete, and that an unresolved change cannot bypass the check. A discussion of what staff would do is not the same evidence as observing what the process permits.

Keep the result with the funding decision, including any route that still needs fixing and who is responsible for it. This demonstrates the behaviour of the paths tested, not that every possible fraud will be prevented.

If the demonstration can still send money to the changed account without the required supplier confirmation, the extra approval has not fixed the problem you were asked to fund.


For FSRA-regulated firms

The Abu Dhabi Global Market Financial Services Regulatory Authority’s GEN 3.5.1 requires a written, governing-body-approved cyber-risk framework with proportionate controls. GEN 3.5.4 also requires assessment of control effectiveness and analysis and quantification of the potential impact of residual cyber risk. These provisions do not prescribe a particular supplier-callback method. The comparison above does not complete that wider assessment.

Incident reporting is a separate obligation. Where GEN 3.5.18 applies, an Authorised Person must notify immediately, and in any event no later than 24 hours, after becoming aware or having information reasonably suggesting that a material Cyber Incident has occurred. Not every supplier fraud automatically meets that definition; other reporting duties may also apply. Obtain qualified advice on the firm’s obligations. A final loss calculation is not a prerequisite for notification.

Sources checked 21 October 2026. The opening and deadline scenarios are illustrative. The NSW incident is an attributed public-sector audit case, not a Cornerstone Red engagement. This article is not regulatory clearance.

Where this applies in our work: Cyber Risk Quantification.