First-Party Fraud Is Rewriting the Loss Curve
First-party fraud — where the person committing fraud is also the account holder — has become one of the fastest-growing loss categories for banks and lenders. Unlike third-party fraud, it's nearly invisible until the losses arrive.
The Fraud Your Models Weren't Built to Catch
Traditional fraud detection was built around a clear adversarial model: a fraudster — someone who is not the account holder — uses stolen credentials or a counterfeit card to make unauthorized transactions. The real cardholder is the victim. The bank's job is to catch the fraudster and make the victim whole.
First-party fraud breaks this model. In first-party fraud, the account holder is the fraudster. They open an account with their real identity, establish a payment relationship, and then exploit that relationship — filing false dispute claims, taking on credit they never intend to repay, or manufacturing transactions to generate refunds or chargebacks. The customer is not a victim. The customer is the threat.
This distinction matters enormously for fraud detection. Most fraud models are built on signals that separate the account holder's behavior from anomalous activity. In first-party fraud, there is no anomaly — the activity is being performed by the person who is supposed to be performing it.
The Scale of the Problem
First-party fraud losses are difficult to measure precisely because they're often categorized as credit losses or operational losses rather than fraud losses — the behavior isn't always recognized as fraudulent until after the fact, if at all. Estimates from industry sources suggest first-party fraud represents 20–40% of total fraud losses at many financial institutions, and the proportion has been growing.
Several dynamics have contributed to the growth:
Bust-out fraud at scale. Bust-out fraud — where a customer builds credit utilization over time, maximizes their credit lines, and then disappears — has existed for decades. What's changed is the sophistication of organized rings that execute bust-out schemes systematically across multiple institutions, coordinating account behavior to avoid triggering any single institution's models.
BNPL and instant credit proliferation. The growth of buy-now-pay-later products and instant credit offers has created more entry points for first-party fraud. Products with light underwriting, fast approval, and digital-only relationships are easier to exploit than traditional credit products with branch relationships.
Dispute abuse normalization. The framing of chargebacks and dispute rights as consumer protections — which they are — has also normalized the idea of disputes as a mechanism for getting money back from merchants for purchases the consumer doesn't want to pay for. "Friendly fraud," where a customer disputes a legitimate transaction, blurs the line between consumer error and intentional fraud, and the volume has increased significantly post-pandemic.
What Detection Actually Requires
Detecting first-party fraud requires different signals than detecting third-party fraud, and it requires different organizational coordination — because it sits at the intersection of fraud, credit, and customer experience in ways that don't fit neatly into any one team's jurisdiction.
Behavioral signals across the relationship lifecycle. First-party fraud often has a behavioral arc — a period of establishing the relationship followed by exploitation. Models that look only at transaction-level anomalies miss the pattern. Models that look at relationship-level behavior over time — how quickly a customer maxes out credit, whether dispute rates spike at certain account ages, how contact information changes track with account behavior — can identify patterns that are invisible in the transaction stream.
Cross-institution data sharing. Bust-out fraud rings operate across multiple institutions simultaneously. No single institution has visibility into the full behavioral pattern. Industry consortium data — shared velocity signals, shared account opening indicators — is one of the few ways to see the pattern that's invisible within a single institution's data.
Calibrated dispute investigation. Not every disputed transaction is fraud — and treating all disputes as potential fraud damages legitimate customer relationships. But treating all disputes as legitimate also creates an invitation to abuse. The calibration between investigation effort and dispute value, informed by risk signals about the disputing customer, is where many programs have room to improve.
Clear ownership of the loss category. First-party fraud losses frequently fall into grey areas between fraud teams, credit risk teams, and operations. When no team clearly owns the problem, nobody is optimizing for the solution. Institutions that have made meaningful progress on first-party fraud have typically assigned clear ownership and built reporting that makes the losses visible as a distinct category.
The loss curve on first-party fraud doesn't improve passively. It requires deliberate model investment, organizational alignment, and a willingness to examine dispute and credit loss patterns with fresh eyes. The institutions doing that work are finding that what looked like credit losses often has a fraud signature that can be detected — and interrupted — before the loss arrives.