AI-assisted fraud costs US financial firms $5.47 per dollar lost

The bill for financial fraud has never been higher. Every dollar of fraud now costs U.S. financial services firms $5.47 in total, up from $4.00 in 2021, as AI-generated attack techniques and a near-total migration of fraud to digital channels push losses to record levels.

That’s according to the LexisNexis Risk Solutions True Cost of Fraud Study 2026 North America, released October 19, 2026, which surveyed 319 risk and fraud professionals across financial services and lending organizations in the United States (251) and Canada (68) and found that websites and mobile apps now account for 69% of all fraud losses.

Nearly all institutions surveyed, approximately 96%, reported that fraud targeting mobile apps increased over the past year, while 69% said mobile fraud costs rose by 10% or more during that period.

Mobile is the new front line

The concentration of fraud in digital channels reflects a broader shift in where financial crime is being waged.

Online portals and websites account for 39% of losses, with mobile apps contributing another 30%. Card transactions and account transfers together represent approximately 48% of losses by payment method, with digital wallets (still a relatively new vector) already accounting for 15%.

Transaction activity is the single most exposed point in the customer journey, attracting 39% of fraud losses across North America. New account creation accounts for 33% and account login or access for 28%, underscoring that the threat follows clients through their entire relationship with a financial institution.

Scams add a particularly complex layer. North American organizations attribute approximately 32% of fraud losses to scams, but U.S. firms reimburse only 31.2% of those losses, compared with 38.9% in Canada, a discrepancy that reflects differing regulatory environments and consumer protection expectations on either side of the border.

Malicious bot attacks compound the picture further, representing 14% of all monthly transactions in the U.S., with 58% of American respondents reporting that bot volume rose year over year.

Earlier this year, FINRA launched a threat-sharing portal for cyber and fraud intelligence to help broker-dealers swap timely information on cybersecurity and fraud threats and coordinate responses.

AI is fighting on both sides

Organizations estimate that fraudsters are using AI-generated techniques in approximately 19% of fraud losses in the U.S. - a figure that rises to 25% in Canada.

That threat has prompted financial institutions to accelerate their own AI deployments. More than half of respondents (approximately 56%) said they are currently using AI to defend against synthetic identity fraud and account takeover, while 48% deploy it against scams.

Synthetic identity fraud - where fraudsters create fictitious identities by combining real and fabricated personal data - is the single largest fraud category, accounting for 19% of losses. Third-party identity fraud, third-party account takeover, and first-party fraud each account for 17%.

Four in five respondents rate AI-powered capabilities as extremely or very important when selecting fraud prevention vendors, and 35% believe advanced AI and machine learning analytics will have the greatest impact on fraud prevention over the next one to two years.

Looking ahead, 62% of institutions plan to invest in AI-enhanced identity verification, 55% in digital identity wallets, and 53% in biometric authentication.

"Fraud continues to evolve across every stage of the customer journey, and digital channels and AI-assisted techniques are increasing the speed, scale and complexity of attacks," said Kimberly White, senior director, fraud and identity, North America, at LexisNexis Risk Solutions, in a statement accompanying the report.

Automation drives better outcomes

The study draws a sharp line between institutions that have built automated fraud programs and those that still rely primarily on manual processes. Organizations with mainly automated fraud programs reported perceived effectiveness of 89%, compared with 61% for manual-first organizations.

They also recorded greater reductions in customer churn - 29% versus 10% - and stronger improvement in overall fraud prevention performance, at 79% versus 54%.

Higher-maturity firms are also more than twice as likely to track fraud by both channel and payment method, a granularity that gives them a meaningful edge in identifying emerging attack patterns. They hired three or more fraud-focused roles in the past two years at a rate of 55%, compared with 29% at manual-first organizations.

That performance gap matters beyond the fraud ledger itself. Approximately 56% of all respondents said their fraud prevention strategies have increased customer churn - a finding that points to a design challenge as much as a security one.

Fraud controls that create excessive friction carry their own revenue cost, and institutions are increasingly measuring success by both metrics. Improving customer trust and loyalty ranked as the top organizational priority for 51% of U.S. respondents, ahead of minimizing fraud losses, which was cited by 43%.

Despite broad consensus on the importance of AI, adoption is uneven. Approximately 31% of North American organizations named limited internal expertise as a top barrier to deploying AI-based fraud detection. Integration complexity with existing fraud and risk infrastructure was cited by 26%, while data quality and availability challenges were flagged by 25%.

The study's methodology covered commercial banks, payment service providers, credit unions, investment and wealth management firms, consumer and auto lenders, mortgage providers, fintechs, and digital banks. Approximately 65% of respondents represented mid-sized and large organizations with annual revenue of $10 million or more.