AI Fraud Prevention: 85% of Merchants Targeted, Industry Demands Better Real-Time Security

AI-Powered Fraud Attacks Surge as Merchants Demand Better Defenses

The fraud landscape has reached critical levels, with three in four European adults encountering scams over the past year according to the Global Anti-Scam Alliance (GASA)1

1

. Research from payabl. reveals that nearly 85% of merchants have been directly targeted by fraudsters over the last 12 months, with 84% now calling for better technology to help them fight back1

1

. Criminal networks are harnessing AI to launch sophisticated attacks at machine speed, forcing financial institutions to rethink their entire approach to payment security.

According to INTERPOL, global fraud losses reached $442 billion in 2025, while Bureau's Global Fraud Intelligence Report 2026 identified approximately 14,000 organized fraud rings in the first half of 2026 alone3

3

. Bureau also detected over 21.7 million account takeover attempts across 5.4 billion login sessions during that period3

3

. These staggering figures underscore how fraudsters see the financial ecosystem as one unified attack surface, while institutions continue defending it as separate companies.

Infrastructure Gaps Undermine AI Fraud Prevention Effectiveness

AI in fraud prevention isn't overhyped, but the idea that organizations can simply purchase an AI-powered fraud tool and solve cybercrime certainly is. Speaking at Fintech Meetup Europe 2026 in Lisbon, industry expert Breno Oliveira emphasized that the problem isn't insufficient AI, but rather inadequate underlying infrastructure to enable effective real-time decision-making1

1

. Many institutions still operate with fragmented systems, static rules, and disparate data sitting silently in different silos.

This technical debt has evolved into operational debt and genuine fraud risk. Legacy systems built over decades store data across multiple platforms with different ownership structures. Changing a policy, integrating a new data source, or deploying a new model can take months, while fraudsters iterate in minutes without change-control committees1

1

. AI-driven fraud detection requires interconnected systems and accessible data encompassing identity, behavior, and transactions. Even the most intelligent machine learning model cannot make effective real-time decisions if relevant signals arrive too late.

Corporate Investment Priorities Shift Toward Prevention Over Detection

PYMNTS Intelligence research surveying 150 U.S. treasury and finance executives at companies with at least $100 million in annual revenue reveals a strategic shift in payment security priorities2

2

. Eighty-one percent of firms plan to adopt or improve automated vendor-onboarding verification during the next 12 months, while 75% cite real-time bank account verification before payment. AI-driven transaction monitoring comes third at 73%2

2

.

This ordering demonstrates where companies see their first line of defense. Rather than relying primarily on technology to identify suspicious transactions after payment instructions enter the process, firms are emphasizing establishing whether vendors and destination accounts are legitimate before money moves. Ranjan R Reddy, Founder and CEO of Bureau, notes that businesses should combine machine learning with device intelligence, behavioral analytics, identity signals, and transaction context to detect anomalies as they occur3

3

.

False Positives Emerge as Critical AI Adoption Barrier

Corporate adoption of AI fraud tools reveals a stark divide: 57% of companies use no AI fraud-detection tools, while 42% use three or more, with just 1% reporting use of only one or two2

2

. This distribution suggests AI fraud prevention is being implemented as a complete stack rather than standalone capability. Once companies cross the adoption threshold, they typically combine applications such as real-time risk scoring, document verification, and scanning communications for AI-generated markers.

However, false positives remain the biggest concern across both adopters and non-adopters. Among firms without AI-driven fraud detection, 89% cite fear of flagging legitimate payments as an adoption barrier, while 74% are hesitant to automate without human sign-off2

2

. Experience with the technology doesn't eliminate these concerns—among AI adopters, 92% say concerns about incorrectly flagging legitimate payments made adoption difficult, and 89% cite hesitation about automating without human approval2

2

.

Reducing Friction While Strengthening Real-Time Fraud Detection

Friction remains a significant challenge in combating real-time fraud. Research shows 43% of merchants believe existing fraud checks slow down the checkout process, while 34% say added friction contributes to abandonment1

1

. The historic response to fraud has often been introducing more friction, but this becomes less effective when machines can churn through that friction indefinitely while legitimate customers grow frustrated.

Adaptive authentication and behavioral analytics offer solutions by applying security measures only where risk signals justify intervention. Businesses should look continuously at identity, behavior, and transaction context throughout the payment journey rather than adding large authentication hurdles at the end1

1

. The goal should be precision—allowing legitimate transactions to pass quietly while intervening only when risk assessment indicates genuine threats. Bureau combines identity, device, and behavioral intelligence to assess activity throughout the customer journey, enabling businesses to apply additional authentication when combined signals indicate elevated risk3

3

.

Network-Level Intelligence Uncovers Coordinated Fraud Operations

AI makes fraud easier to personalize, automate, and scale. Attackers can combine exposed customer information with data from other sources to craft convincing phishing campaigns, impersonate legitimate users, and target enterprise platforms3

3

. AI-generated content makes these attempts harder for conventional security systems to distinguish from genuine activity, particularly when stolen personal information appears legitimate.

Machine learning can detect recurring patterns across devices, identities and transactions, while network analysis maps connections that individual account checks may overlook. Multiple accounts may have different owners and transaction histories yet share devices or other indicators linked to suspicious activity3

3

. Bureau's Graph Identity Network (GIN) analyzes these relationships to support real-time decision-making, identifying the entities linking coordinated operations. Network-level intelligence adds crucial dimensionality by connecting suspicious activity across accounts, devices and platforms, helping institutions identify coordinated operations that transaction-level monitoring misses.

Industry Gaps and Payment Method Vulnerabilities Persist

Industry comparisons expose particularly sharp divides among companies selling physical goods. Just 34% of goods-sector firms use three or more AI fraud-detection tools, compared with 44% of services companies and 57% of technology firms2

2

. Goods companies also have the lowest share stopping at least 90% of fraud attempts before a loss, at 41%2

2

.

The AI discussion can obscure older sources of payment risk like checks. Thirty-five percent of surveyed firms report encountering check fraud, with 48% saying checks account for at least 10% of outbound fraud losses, including 15% that attribute more than one-quarter of their losses to checks2

2

. AI-powered detection can improve monitoring around a payment, but the payment method itself remains part of the risk equation. This gives payment modernization a security dimension alongside its familiar speed, cost and automation benefits, with changing how money leaves the business sitting alongside improving the technology watching it.