Stripe Capital's 81,000 MCAs and the Broker-to-Funder Fraud Gap Nobody Talks About

Stripe's 81,000 MCAs Reveal a Structural Advantage Most Funders Can't Replicate

When deBanked reported that Stripe Capital originated 81,000 merchant cash advances and business loans in 2025, the headline number grabbed attention. But the real story is what sits underneath it: Stripe never has to ask a merchant for a bank statement. It never schedules a verification call. It never worries about whether a broker doctored a PDF before forwarding it to the funding desk. Stripe already sees the merchant's revenue flowing through its own payment rails. AI fraud detection for business lending, in Stripe's case, starts with data the platform already owns.

That structural advantage is worth understanding, not because every funder can replicate it, but because it exposes the exact vulnerability that independent MCA lenders need to solve. If you fund merchants who come through ISO channels or broker networks, you are operating with secondhand data. The question is whether your verification process is strong enough to compensate for the information gap. In 2026, the answer for too many funders is still no.

This article breaks down the fraud exposure created by the broker-to-funder handoff, explains how AI-powered verification is evolving to address it, and offers a practical framework for independent lenders who want platform-level confidence without platform-level data access.

The Broker-to-Funder Handoff: Where Fraud Enters the Pipeline

Why the Handoff Creates Vulnerability

Platform lenders like Stripe and Square have a closed-loop advantage. They process the merchant's card transactions, so they can underwrite directly from verified revenue data. Independent funders, by contrast, rely on a chain of custody that introduces risk at every step. A merchant applies through a broker. The broker collects bank statements, sometimes a voided check, perhaps a driver's license. Those documents get forwarded to the funder, sometimes through a CRM, sometimes as email attachments.

At every transition point, the document can be altered. Bank statements can be edited with consumer-grade PDF tools. Transaction histories can be fabricated. Account balances can be inflated. The funder receiving these documents has no way to confirm they came directly from the bank unless they verify independently. This is the broker-to-funder fraud gap, and it persists because many funders still treat document collection as a trust exercise.

The Scale of the Problem

The FBI's prosecution of Saul Shalev, recently detailed by deBanked, underscores just how lucrative MCA fraud can be. Prosecutors allege Shalev funded a lavish lifestyle through systematic fraud against small business finance companies. Cases like this are not anomalies. They represent the predictable outcome of a system where document authenticity is assumed rather than proven. As we explored in our analysis of the Carroting scam and the FBI's latest MCA fraud case, organized fraud rings specifically target the verification gaps in the broker-to-funder pipeline because they know where the weak points are.

The fundamental issue is simple: static documents are easy to forge. A PDF bank statement is a snapshot that can be reconstructed pixel by pixel. Even OCR-based verification tools that scan for font inconsistencies or metadata anomalies are playing catch-up against increasingly sophisticated editing techniques.

How Platform Lenders Bypass the Problem Entirely

Stripe does not verify bank statements because it does not need to. Its underwriting model ingests the merchant's actual payment processing data, including transaction volume, refund rates, chargeback frequency, and seasonal patterns. Square operates similarly, which is why it reported $7 billion in lending in 2025 with virtually zero nonperforming loans. Their AI models are trained on first-party data that has never passed through a broker's hands.

This creates a two-tier market. Platform lenders operate with high-fidelity data and can automate decisioning at scale. Independent funders operate with lower-fidelity data and must compensate with more manual verification, which costs time and money and still leaves gaps.

AI Fraud Detection Is Moving Beyond Document Scanning

The Limits of OCR-Based Verification

Most AI fraud detection for business lending today focuses on document analysis. OCR engines extract data from bank statements, and machine learning classifiers flag anomalies: mismatched fonts, inconsistent spacing, metadata that suggests the file was edited in Adobe Illustrator rather than exported from a banking portal. These tools have value. They catch unsophisticated fraud and automate the tedious work of manual statement review.

But they have a ceiling. A skilled forger using the right tools can produce a bank statement that passes OCR validation. The fonts will match. The spacing will be consistent. The metadata will look clean. Document-level AI is necessary but not sufficient.

Behavioral Verification: The Next Frontier

The more robust approach is behavioral verification: instead of analyzing a static document, verify the behavior of the applicant interacting with their live banking environment. This means watching a merchant log into their actual bank portal, navigate to the account summary, scroll through transaction history, and display the balances that matter for underwriting.

This is fundamentally harder to fake. A fraudster can edit a PDF in minutes. Recreating a fully functional banking portal with consistent transaction data, working navigation, and real-time page loads is orders of magnitude more difficult. The verification shifts from "does this document look real?" to "is this person actually showing me a real banking session?"

Exact Balance was built around this principle. Rather than asking applicants to upload static documents, the platform sends them a secure link to record their browser-based banking session. An AI-guided floating coach walks them through each step, verifying completion in real time. The resulting recording captures the live banking portal, including URL bar, page transitions, account details, and transaction data, all timestamped and stored securely on encrypted cloud infrastructure.

AI Step Detection and Recording Validation

Where AI adds the most value in this model is not in reading documents but in validating behavior. Exact Balance's AI vision analyzes the recording to confirm that the applicant completed each required step: logging in, displaying the account summary, navigating to the correct date range, and showing specific transaction details. The system can flag recordings where steps were skipped, where the URL does not match a recognized banking domain, or where the session exhibits anomalies like suspiciously fast navigation that might indicate a pre-recorded or scripted session.

This layer of AI fraud detection for business lending operates at the session level rather than the document level. It does not replace human review; rather, it triages recordings so that underwriters can focus their attention on cases that warrant closer inspection. The combination of AI-powered validation and human judgment creates a verification standard that approaches the data confidence platform lenders enjoy, without requiring access to the merchant's payment processing rails.

A Practical Framework for Independent Funders

If you are an independent MCA funder competing against platforms with native data access, here is how to close the verification gap without rebuilding your entire technology stack.

Stop treating static bank statements as ground truth. PDFs should be a starting point, not the finish line. Layer live verification on top of every deal where the funding amount justifies the extra step. For most funders, that means every deal.

Eliminate the scheduling bottleneck. One of the reasons funders skip thorough verification is that live calls are painful to schedule. Applicants are busy. Time zones create friction. Calls get rescheduled, and deal velocity suffers. Asynchronous verification solves this by letting the applicant record on their own time. As we detailed in our comparison of screen recording versus live verification calls, the async model consistently delivers faster turnaround with higher completion rates.

Use AI to triage, not to decide. The most effective AI implementations in MCA underwriting augment human judgment rather than replace it. Let machine learning flag anomalies, validate recording steps, and surface the cases that need closer review. But keep a human underwriter making the final call. Regulators are watching how AI is used in credit decisions, and the Consumer Financial Protection Bureau has signaled increasing scrutiny of automated decision-making. Having a clear human-in-the-loop process protects you legally and produces better outcomes.

Build a complete audit trail. Every verification should produce a timestamped, securely stored artifact that you can produce for regulators, legal disputes, or internal review. Screen recordings with activity logs accomplish this naturally. A recording of a live banking session is far more defensible than a PDF that may or may not have been altered before it reached your desk.

Independent funders will never have Stripe's data advantage. But with the right verification workflow, they can achieve comparable confidence in the data they use to underwrite. The gap is not about technology access; it is about process design.

Frequently Asked Questions

How does AI detect fraud in MCA lending?

AI detects fraud in MCA lending through multiple layers: document-level analysis using OCR and machine learning to flag inconsistencies in bank statements, behavioral verification that analyzes screen recordings of live banking sessions for anomalies, and pattern detection across application data to identify stacking or synthetic identity fraud. The most effective approaches combine automated AI triage with human underwriter review, using AI to surface suspicious cases rather than make final decisions independently.

What is broker-to-funder fraud in the MCA industry?

Broker-to-funder fraud occurs when documents are altered or fabricated during the handoff between a broker who collects merchant applications and the funder who provides capital. Because bank statements, identification documents, and financial records pass through intermediaries before reaching the underwriting desk, each transition point creates an opportunity for tampering. This is one of the most common and least discussed fraud vectors in merchant cash advance lending.

Can screen recordings replace traditional bank statement verification?

Screen recordings of live banking sessions provide stronger verification than static bank statements because they capture real-time interaction with an authenticated banking portal. They do not fully replace document review but serve as a superior primary verification method. The recording shows the actual banking environment, including URLs, page navigation, live balances, and transaction histories, making it exponentially harder to falsify compared to a PDF document.

How do platform lenders like Stripe verify merchant revenue without bank statements?

Platform lenders verify merchant revenue by analyzing first-party transaction data that flows through their own payment processing infrastructure. Because they handle the merchant's card transactions directly, they have real-time visibility into sales volume, refund patterns, and cash flow consistency without needing to request external bank statements. Independent funders who do not process the merchant's payments must use alternative verification methods to achieve comparable data confidence.

Conclusion

Stripe's 81,000 originations in 2025 highlight a widening gap between platform lenders with native data access and independent funders who depend on broker-sourced documents. The broker-to-funder handoff remains the industry's most exploitable fraud vector, and static bank statements alone are no longer sufficient to mitigate the risk.

AI fraud detection for business lending is evolving from document scanning to behavioral verification: analyzing how applicants interact with their live banking portals rather than trusting the documents they submit. Independent funders who adopt async, AI-guided bank verification can close the data confidence gap without needing to own the merchant's payment rails.

Exact Balance was built for exactly this challenge. Visit exactbalance.ca to see how asynchronous screen recording with AI-guided validation fits into your underwriting workflow, and start verifying with the same confidence that platform lenders take for granted.