Prevent AP and AR Fraud: A Playbook to Stop AI-Fueled Attacks

To prevent AP and AR fraud, finance teams need layered controls: strict segregation of duties in the vendor master file, positive pay and ACH debit blocks at the bank, dual approval for payment changes, and out-of-band verification for any banking detail update. No single control is enough on its own, especially now that AI-generated emails and deepfake voice calls can convincingly impersonate your CFO or a longtime supplier.

Why the Threat Has Escalated

The numbers tell a stark story. According to the 2026 AFP Payments Fraud and Control Survey, which reports on 2025 activity, 76% of U.S. organizations experienced attempted or actual payments fraud in 2025. Business email compromise (BEC) affected 74% of organizations, and check fraud struck 58%. The FBI’s 2025 Internet Crime Report logged 22,364 AI-related fraud complaints with losses of nearly $893 million, and BEC as a whole generated about $3.04 billion in reported losses.

The newer and more alarming development is the AI layer being grafted onto these schemes. Fraudsters now use voice-cloning tools to call a controller in the CFO’s voice, reinforcing a wire-transfer request that arrived by email minutes earlier. In one widely documented case, a finance employee at engineering firm Arup transferred roughly $25 million across 15 transactions after a video conference in which every participant, including the apparent CFO, was an AI-generated deepfake. Industry security researchers estimate that AI-assisted deepfake involvement in BEC has climbed from under 5% of incidents in 2023 to around 40% by early 2026, though these are vendor estimates rather than official government figures.

AI also works against fraud, but only if the underlying process controls are already sound. Anomaly-detection tools can flag unusual payment patterns, but they cannot substitute for the foundational controls described below.

The Core Attack Surfaces in AP and AR

Before building a control framework, it helps to map where the attacks actually land.

Vendor Master File Manipulation

The vendor master file is the highest-value target in AP. An attacker who can insert or modify a vendor’s banking details will receive every subsequent payment to that supplier until someone notices. In AFP survey data, vendor impersonation ranks among the most common BEC tactics, cited by a majority of affected organizations, and it has been one of the fastest-growing variants in recent years. The attack pattern is almost always the same. An email appears to come from a known supplier requesting a bank account update, and AP processes it because there is no friction in the workflow.

Invoice Fraud and Duplicate Payments

Ghost invoices, inflated invoices, and duplicates exploit gaps in three-way matching and approval thresholds. An employee who controls both vendor setup and payment approval can create fictitious vendors and approve their own invoices. Duplicate payments often originate from rushed month-end processing where volume overwhelms manual review.

ACH and Wire Fraud

ACH credit-push fraud has become a primary target because ACH transactions can be difficult to reverse once they settle. Nacha’s credit-push fraud monitoring rules, which rolled out in two phases in 2026 (Phase 1 effective March 20, Phase 2 effective June 19/22), now require all non-consumer originators, third-party senders, and third-party service providers to implement risk-based fraud monitoring regardless of transaction volume. For AP teams, this means velocity checks, anomaly detection, and behavioral-tolerance monitoring are now a compliance baseline, not merely a best practice.

AR Skimming and Remittance Fraud

Fraud also runs through the revenue side. An employee with access to incoming payments can redirect checks or ACH credits before they post. On the customer-facing side, fraudsters sometimes impersonate a company to redirect customer payments, a variant called AR hijacking that is difficult to detect until customers follow up on overdue balances.

Controls That Prevent AP and AR Fraud

Positive Pay for Checks and ACH

Positive pay is the single most effective tool for stopping check fraud. Your treasury team sends the bank a daily file listing every check issued: check number, amount, payee name, and date. The bank compares each presented item against that file and flags exceptions for your approval before clearing. Payee positive pay, the version that verifies the payee name, provides stronger protection than amount-only matching. Without positive pay, counterfeit and altered checks can clear undetected.

ACH positive pay (sometimes called ACH debit filter or ACH debit block) works the same way on the electronic side. You maintain a list of authorized ACH debit originators, and any debit outside that list is blocked or returned. Combined with ACH credit-push fraud monitoring now required under the 2026 Nacha rules, this creates a meaningful barrier on both the debit and credit sides of the ACH channel.

Dual Approval for Payment Changes

No single person should be able to make a payment and also change the vendor record it flows to. Dual approval means two separate, independent employees must authorize any of the following before it takes effect:

• New vendor additions to the master file
• Changes to vendor banking or remittance details
• Wire transfer initiation above a defined threshold
• ACH credits above a defined threshold or to a new account

The second approver must be entirely independent, meaning they have no financial interest in the transaction and did not initiate it. Many organizations require that approval for banking-detail changes come from a supervisor at least one level above the requestor.

Out-of-Band Verification for Banking Changes

Dual approval alone is not enough when sophisticated attackers can also compromise a second employee’s email. Any change to a vendor’s banking details should trigger a callback to a phone number already on file, specifically one retrieved from your own system, never from the change request itself. The call must confirm the specific change, not just that the vendor exists. Document the verification, including the name of the person reached, the date, and what was confirmed.

This callback protocol is the single highest-impact control for stopping vendor-impersonation BEC. Organizations that enforce it essentially close the most common attack vector.

Segregation of Duties in the Vendor Master

Three roles must be separated to maintain a defensible vendor master file: (1) creating or modifying vendor records, (2) approving invoices, and (3) initiating or approving payments. If any two of these roles sit with the same person, the risk of undetected fraud is materially elevated. Smaller finance teams often struggle here, and the right answer is compensating controls: management review of all new vendors and recent-change vendors on a defined cadence, automated system flags on same-person vendor-creation and payment-approval combinations, and periodic vendor master audits.

Three-Way Match and Invoice Controls

Three-way match, comparing a purchase order, a receiving document, and the invoice, prevents payment for goods or services never ordered or received. It is foundational, but it only works if the underlying PO and receiving data are accurate and independently created. Configuring your ERP to require three-way match clearance before payment release, and reviewing open exceptions rather than batch-clearing them, makes the control meaningful rather than nominal.

Set and enforce approval thresholds. Invoices above a threshold should require a second approval. Establish exception reporting for: invoices with no PO, invoices that bypass standard match tolerance, vendors with recent address or banking changes, and any payment made to a vendor added in the prior 90 days.

Building the AI-Aware Control Layer

AI-enabled fraud demands an AI-aware response, but that does not mean replacing process controls with software. The correct model is to layer detection capabilities on top of strong foundational controls.

Account Validation at Onboarding

Before making a first payment to any new vendor, validate that the routing number and account number actually belong to an entity consistent with the vendor’s name. Bank account validation services cross-check account ownership in real time. This is increasingly required under the 2026 Nacha framework and is a practical backstop against synthetic vendor fraud.

Behavioral Anomaly Detection

Transactions that deviate from established patterns, a vendor that has never received more than $50,000 per month suddenly getting a $300,000 wire, a payment to a vendor created 3 days ago, a wire to an international account where the vendor has always been domestic, deserve a review hold, not automatic processing. Many ERP and AP automation platforms now include configurable anomaly detection that can flag these patterns for human review before the payment releases.

Deepfake and Impersonation Response Protocols

Train finance staff to recognize that any out-of-band request, even a phone call or video conference, is not automatically trustworthy. Establish a code word or challenge-phrase protocol for urgent payment requests from executives. Any request that bypasses normal approval workflows, regardless of how it arrives or how convincing the requestor seems, should trigger the standard control sequence rather than override it. Speed and urgency are tactics attackers use deliberately, and your controls need to be explicitly resistant to time pressure.

AR-Side Controls

Accounts receivable fraud is less discussed but equally costly. Controls on the AR side should include:

• Segregation between cash application (posting payments) and customer account maintenance (credit adjustments, write-offs)
• Independent reconciliation of the AR subledger to the general ledger, performed by someone who does not post cash
• Lockbox or electronic remittance arrangements that route payments directly to the bank rather than through individual employees
• Prompt follow-up on customer accounts that appear current in the subledger but where the customer disputes outstanding balances, a potential indicator of AR skimming

How Modus Approaches AP and AR Risk

When Modus performs audit and assurance engagements or advisory reviews of finance operations, AP and AR controls are evaluated as a connected system, not a checklist. AI-native audit workflows allow the team to analyze full populations of transactions, flagging anomalies across all vendor payments rather than relying on sampling alone. That means patterns that sample-based testing would miss, like a fictitious vendor receiving small payments over 18 months, surface during fieldwork rather than after a loss.

For companies that want a proactive assessment rather than waiting for an audit cycle, a targeted internal controls review scoped to AP and AR can identify gaps in segregation, approval workflows, and banking-change procedures before a loss occurs.

Related reading: Finance Transformation with Cloud ERP: How to Sequence It Right

Frequently Asked Questions

How do I prevent AP fraud?

Preventing AP fraud requires layered controls: strict segregation of duties between vendor master maintenance, invoice approval, and payment release; positive pay for checks and ACH debit blocks for electronic payments; dual approval with out-of-band callback verification for any vendor banking change; and three-way match enforced at the system level. No single control is sufficient on its own.

What is positive pay and how does it prevent fraud?

Positive pay is a bank service where your company submits a daily file of every check or ACH debit it authorizes. The bank matches incoming items against that file and holds exceptions for your review before clearing. It is highly effective against counterfeit and altered checks and unauthorized ACH debits. Payee positive pay, which also verifies the payee name, provides stronger protection than amount-only matching.

What is dual approval in accounts payable?

Dual approval requires that two separate, independent employees authorize a transaction or change before it is processed. In AP, it most commonly applies to vendor banking-detail changes, new vendor additions, and payments above a defined threshold. The second approver must have no role in initiating the request and no financial interest in the transaction.

How does business email compromise target AP departments?

BEC attackers impersonate a known vendor or executive via email, requesting a change to banking details or an urgent wire transfer. Once the change is processed, subsequent legitimate payments flow to the attacker’s account. AI-generated emails and voice clones now make these impersonations nearly indistinguishable from genuine communications without a verified out-of-band callback to a phone number already on file.

What are the 2026 Nacha ACH fraud monitoring requirements?

Nacha’s credit-push fraud monitoring rules require all non-consumer ACH originators, third-party senders, third-party service providers, and ODFIs to implement risk-based fraud monitoring. Phase 1 took effect March 20, 2026 for larger-volume participants, and Phase 2 removes the volume threshold for the remaining participants effective June 19, 2026 (with the next banking day, June 22, 2026, as the practical date). Acceptable methods include velocity checks, behavioral-tolerance monitoring, anomaly detection, and pattern recognition. On the receiving side, RDFIs must also monitor for incoming ACH credit entries that appear to be initiated due to fraud.

How do I protect the vendor master file from fraud?

Separate the roles of vendor record creation, invoice approval, and payment release so no individual controls all three. Require a company-supplied verification form, including tax ID and prior remittance address, for any banking-detail change, then call a number from your own records, never from the change request, to confirm. Flag and review all vendors modified in the prior 90 days before releasing payments. Audit the vendor master periodically for duplicate bank accounts, missing tax IDs, and vendors with no recent transaction history.