Fraud prevention costs outpace fraud losses
Banks now spend more than $2 to stop every $1 lost to fraud.
Financial institutions are effectively on a treadmill, running faster just to stay in place. In 2026, the fraud loss rate (i.e., the share of transaction dollars lost to fraud) averaged 0.58 basis points. One basis point equals 0.01%, so that works out to about 58 cents lost for every $10,000 moved. That figure is almost identical to the 0.59 basis points financial institutions reported in 2024 after a jump to 0.81 in 2025. Estimated losses per institution reached $1.92 million in 2026, up 8% from $1.78 million in 2024.
That’s after extensive investment in fraud prevention. Spending on staff, software and outsourced fraud services climbed from $1.67 million per institution in 2024 to $4.35 million in 2026. Prevention now takes up 69% of the total cost of fighting fraud, up from 49% two years ago. So, financial institutions now spend $2.27 to stop every $1 that still gets through, more than double the $0.94 they spent in 2024.
The total cost of fraud, losses plus prevention combined, nearly doubled year over year, climbing to $6.27 million per institution from $3.39 million in 2025. Nearly nine in 10 financial institutions (88%) increased their fraud budgets in that time, and none cut spending.
Fraud risk is shifting to new players
Large banks have built up strong defenses, but neobanks and FinTechs are struggling.
It’s not as if banks’ efforts are useless. Large banks, the ones with the biggest budgets and the longest-standing fraud programs, cut their losses nearly in half, from $16.78 million per institution in 2025 to $9.00 million in 2026. Their fraud loss rate dropped 68% to 1.13 basis points from 3.58. Spending at these institutions held roughly flat.
Neobanks and FinTechs, meanwhile, are losing the battle. Their losses more than tripled in the last year, increasing from $0.80 million to $2.84 million per institution, while their spending to fight fraud more than quadrupled from $1.75 million to $8.02 million. That’s roughly four and a half times the $1.76 million a typical regional or community bank now spends. For those local banks, losses increased only a small amount, from $0.95 million to $1.05 million, while spending rose from $1.52 million to $1.76 million.
Fraud, in other words, is going where there’s growth. The institutions signing up the newest customers and running the fastest payment rails are the ones now highly vulnerable to fraud, while the largest, most established banks have built up strong defenses.
Fraud is getting past defenses
Banks have built strong defenses—but fraud is coming through the front door.
Financial institutions have been very successful in building defenses against break-ins from the outside. Behavioral analytics, device fingerprinting and enhanced knowledge-based authentication (KBA, a method of verifying identity through security questions and personal details) each reached 100% adoption in 2026, up from 81%, 95% and 68%, respectively, in 2025. Additionally, 96% of institutions have machine learning and artificial intelligence (AI) tools, and just as many have implemented rules-based algorithms. KBA showed the biggest rise in adoption, up 47% in a single year.
Now, the fastest-growing fraud doesn’t trip those wires. It walks right in using the customer’s own login or the customer’s own consent. PYMNTS Intelligence sorts fraud using the Federal Reserve’s FraudClassifier model, a framework that groups fraud by who started the payment, what went wrong and how the fraudster got in, so that different institutions can compare fraud types on equal footing.
Under that framework, fraud where the account holder was manipulated into acting, commonly known as a scam, rose to 22.4% of fraud dollars in 2026, from 12.0% in 2025. Relationship or trust fraud, where a fraudster builds a personal rapport with the victim first, more than doubled to 12.6% of dollar value, up from 6%. Account takeover through compromised credentials grew 82%, to 28.2% of fraud dollars. In total, more than half of all fraud dollars (50.6%) now move through a legitimate login or involve manipulating a customer, up from 27.6% two years ago.
The fraud problem is shifting faster than the technology stack.
Less fraud is coming from the channels that banks’ current defenses were built to catch. Physical forgery and counterfeit fraud fell 66% to 6.1% of dollar value, and impersonation of the account holder fell 37% to 10.8%. Taken together, these older, more visible fraud types dropped from 35% of dollar value to 17%, a 52% decline.
This shift shows up as friction for financial institutions. Seventy-eight percent of institutions now cite the complexity of new technology systems as their top barrier to innovation, up from 48% a year earlier, the single largest jump among the 14 barriers PYMNTS Intelligence tracks. In effect, widespread tool adoption solved one problem and created another: Institutions are stacking overlapping systems on top of each other, and integrating them well has become the new challenge.
Fraud as a trust problem
Scams are increasingly eroding customer trust.
More than just being a number on a loss statement, fraud is hurting the customer relationship. Banks now cite damage to customer loyalty as the most common consequence of fraud, reported by 69% of financial institutions, up from 50% in 2025. Damage to brand and reputation follows close behind at 65%, up from 44%. Both now outrank financial losses, which 55% of institutions experienced.
Almost every other consequence PYMNTS Intelligence tracks got worse, too. Operational disruptions rose to 63% from 47%. Lowered employee morale jumped to 54% of institutions from 34%, the fastest-growing problem. Regulatory action followed close behind, up to 48% of financial institutions overall, and 69% at large banks specifically. Only one outcome improved: Lost new-business opportunities dropped to 45% from 48%, suggesting fraud costs financial institutions more in lost customers than in customers they’re chasing.
New problems demand new solutions. Now, improved communication with customers, not new detection models, is the top investment planned for the next year, chosen by 78% of institutions, up from 41% in 2025. When the biggest risk moves from the transaction to the relationship, the response has to follow the same path.
Banks want control, but not full independence
Financial institutions want to control their fraud defenses, but they’re not building the technology from scratch.
Financial institutions want their own fraud detection tools. In 2026, 71% of institutions reported plans to develop new in-house systems, up from 44% the year before. At the same time, plans to outsource detection and prevention to a third party fell to 29% from 51%. Large banks remain the most willing to outsource (55% plan to), while regional and community banks are the least likely (20%).
Those plans to build in-house are not as independent as they may sound. Among financial institutions developing systems in-house, 93% say that work still combines third-party fraud technology with their own proprietary tools. Only 0.2% describe their in-house work as fully their own, built and run entirely by their own team. Of the 11 fraud technologies PYMNTS Intelligence tracks, no institution reported building any of them entirely independently. Financial institutions aren’t moving away from relying on vendors, but they are looking for ways to gain more control.
That approach may be paying off. Among institutions using machine learning, 86% now say doing so has allowed them to be proactive, catching problems before they happen, up from 25% in 2025. Now, large banks (93%), neobanks and FinTechs (91%) and regional and community banks (83%) all tend to describe machine learning’s effect as mostly proactive rather than reactive.
Read More
PYMNTS Intelligence is the leading provider of information on the consumer trends driving innovation in consumer finance, digital payments and financial inclusion. To stay up to date, subscribe to our newsletters and read our in-depth reports.
Methodology
“Scams Get Personal: Why Banks Pay More to Fight Fraud” is a PYMNTS Intelligence report based on a survey of 90 executives at U.S. banks, credit unions and FinTechs, fielded in July 2026. Eligible banks and credit unions held at least $1 billion in assets, and eligible FinTechs reported at least $100 million in annual revenue. Findings are compared against two earlier surveys: 200 respondents fielded in October 2025 and 200 respondents fielded in September 2024.
