AI is helping banks save millions by transforming payment fraud prevention

• On average, organizations lost $60 million globally to payment fraud in a year, according to 2025 research by Mastercard. 
• Impersonation scams are rising quickly, fueled by generative AI (gen AI) and resulting in fraud.
• AI tools that use real-time data and trends can help stop fraud before it happens — and reduce losses.
• 83% of industry leaders say AI has reduced false positives and churn, marking a new era in fraud prevention.

The expansion of real-time payment networks has created new opportunities for fraud. Transactions are processed within seconds, making the window for detection and deflection smaller. Meanwhile, criminals design attacks that move money quickly through multiple accounts before getting caught.

Those fraudsters use gen AI to run social engineering scams and organize rapid transactions with more precision. According to a report from Deloitte, gen AI could fuel $40 billion in U.S. fraud losses by 2027 — over triple the $12.3 billion seen in 2023. By automating the flow of funds across accounts in different institutions, fraudsters can obscure the trail and complicate recovery.

But banks are stepping up, too — using AI to fight back faster. In the past, fraud prevention teams set manual rules to decide which transactions to approve or block. Now, AI can analyze data to detect unusual patterns and make smarter authorization decisions in real time.

This shift is already paying off. Mastercard’s 2025 payment fraud prevention report, produced in partnership with Financial Times Longitude, found that 42% of issuers and 26% of acquirers ​​​​​​​​have saved more than $5 million in fraud attempts over the past two years thanks to AI. But to make the most of AI for fraud prevention, organizations need high-quality data that feeds into models for better risk decisions.

Mastercard’s research found that organizations lost an average of $60 million to payment fraud in the past year. And as AI helps fraudsters create more convincing scams, those losses are expected to continue rising:

• Criminals harness AI to analyze large volumes of publicly available data and craft highly targeted scams.
• With affordable gen AI tools, fraudsters can create convincing phishing messages and fake storefronts, establish shell or ghost merchant accounts, and generate synthetic voice clones and deepfake videos at scale, making it easier to manipulate victims into approving fraudulent transactions.

At the same time, as real-time payments gain popularity, financial institutions have only a short window to identify and block AI-powered fraud attempts. That’s why payment industry leaders are keeping a close eye on a variety of fraud risks:

• Leaders see synthetic identity fraud (61%), impersonation scams (60%) and cross-border fraud (54%) as the fastest-growing threats over the next year. Synthetic identity fraud, which involves combining stolen personal information with fabricated details to create a fake identity, is escalating as fraudsters use AI to sift through massive datasets and build more convincing profiles.
• Other growing risks include e-commerce fraud (47%), Buy Now, Pay Later (BNPL) abuse (42%) and deepfakes (21%). Concerns about e-commerce fraud and BNPL abuse, when bad actors make unauthorized BNPL accounts or transactions, underscore the need for stronger customer protections across the digital payments landscape.

Ninety percent of payment leaders expect higher financial losses in the next three years if they don’t increase their use of AI in fraud prevention.

Fortunately, there are various ways to apply AI for fraud prevention, from analyzing transaction patterns to reducing manual reviews. Many institutions are already generating high return on investment (ROI) as a result:

• 85% of respondents report seeing returns from using AI for fraud case triage and investigation, transaction pattern recognition and real-time detection of suspicious transactions.
• 83% say AI has significantly sped up their process for fraud investigation and case resolution.

​​​Still, it’s important to remember that sustained investment leads to the biggest gains. Organizations that have used AI for over five years report saving $4.3 million in lost revenue, almost double the average savings of $2.2 million.

Simultaneously, leaders know that keeping AI current as fraud tactics evolve is a significant hurdle. To defend against new and emerging threats, AI tools need to learn and adapt in real time.

With AI tools that incorporate near real-time data and contextual insights, organizations can make more efficient authorization decisions to increase approval rates and keep customers happy.

The problem: Historically, banks have used rules-based systems to approve or block transactions. For example, a bank may set a rule to flag purchases over a certain dollar amount or deny transactions originating from unusual locations. But here’s the issue: Rigid, manual processes can slow fraud detection, especially as scams get faster and more complex. Issuers and acquirers don’t have equal visibility across each other’s portfolios, creating gaps as fraud targets every stage of the payment lifecycle, including, increasingly, the merchant level, where relevant risk signals may sit outside an individual issuer’s view.

The solution: AI solutions can easily overcome that limitation, analyzing millions of data points to rapidly evaluate transaction risk and offer real-time insights. This means banks can spot emerging threats as they unfold and quickly make informed decisions, minimizing detection lag.

It’s no surprise that 80% of organizations reported that AI helped eliminate unnecessary manual reviews. This potential becomes even more powerful when AI is paired with Mastercard’s global network intelligence to generate risk scores and predictive insights that help fight fraud. Mastercard’s Decision Intelligence Pro solution (DI Pro) helps with exactly that: it provides a risk score and insights in near real time during authorization.

Complementing DI Pro, Mastercard’s Merchant Scam & Risk Indicator (MSRI) is a new AI-driven tool that delivers a predictive merchant-level risk score and reason code during authorization as well, helping issuers identify merchants with elevated scam and fraud-related risk sooner. Together, these insights can help banks make more informed decisions during authorization, focusing manual review where it is needed most.

The problem: Because static authorization rules lack nuance, they often result in false positives, i.e., when legitimate transactions are incorrectly identified as fraudulent and declined or flagged for review. Beyond creating extra work for internal teams, this also hurts the customer experience. 

The solution: Advanced AI models can analyze diverse data points to assess fraud risk with precision. For example, if a customer who regularly buys mid-range clothing suddenly purchases several luxury fashion items during a seasonal sale, advanced AI models can analyze factors like historical purchase information, merchant credibility and timing to determine if this spike in spending is legitimate.

This contextual insight drives more accurate authorization decisions that prevent fraud without increasing friction for customers. In fact, 83% of respondents report that AI has significantly reduced false positives and customer churn rates in the past year. 

The problem: While AI can support real-time fraud detection, it requires high-quality data to remain effective. Leaders understand this demand, with 64% of respondents saying they need to accelerate access to new, credible data sources to keep pace with evolving threats. Forty-one percent consider watchlist and blocklist data among the most important datasets for achieving their future fraud prevention goals, and almost a third say the same of merchant data.

The solution: Effective AI fraud detection models integrate inputs from across the payment ecosystem, including card network intelligence, merchant-level risk indicators and insights from consumers’ digital identities.

For example, an AI tool might assess a merchant’s risk based on historical fraud rates while also analyzing a customer’s velocity count, which tracks how frequently actions like purchases or account changes occur within a short period.

Using advanced AI and the Mastercard Merchant User Connected Intelligence (MUCI) algorithm, DI Pro connects patterns across users, merchants and transactions to provide context, not just a score. MSRI uses a graph neural network (GNN) model to identify merchants with elevated risk, drawing on relationships across merchants, cards, breach clusters and compromised entities and informed by Mastercard’s network and cyber threat intelligence.

Going forward, organizations’ success with AI fraud detection will depend on two factors: a model’s capacity to analyze large volumes of data and its ability to combine historical patterns and fresh information to assist with decision-making.

Generative AI is changing the fraud landscape. Fraudsters are faster and more adaptive with AI — and financial institutions must be, too.

Mastercard’s Decision Intelligence Pro solution uses AI and network-driven insights to analyze and score transactions based on the level of risk. Complementing these transaction-level insights, MSRI uses three predictive AI models to assess distinct dimensions of merchant risk, delivering a single actionable merchant risk score at authorization. With richer insights, you can make confident authorization decisions, approve more genuine transactions and protect revenue as fraud tactics continue to evolve.

Ready to transform your fraud defenses with AI? Explore how Decision Intelligence Pro and Merchant Scam & Risk Indicator can help you identify fraud risk earlier and protect customers.