Fraud Prevention in 2026: 3DS2, AI Scoring, and the Liability Shift
Card-not-present fraud hit record levels in 2025 and continues to climb in 2026. 3D Secure 2.0, device fingerprinting, velocity rules, and AI-based risk scoring form the modern fraud prevention stack, and together they eliminate the vast majority of fraudulent transactions before they authorize.
Card-not-present (CNP) fraud, transactions where neither the card nor the cardholder is physically present, represents more than 80% of all payment card fraud in 2026. As EMV chip technology made in-person terminal fraud increasingly difficult, fraudsters shifted decisively to online channels and virtual terminal/phone orders. In 2025, global CNP fraud losses exceeded $34 billion, and AI-generated synthetic identities and deepfake-assisted account takeovers are accelerating the trend in 2026. Roots' fraud prevention suite addresses every layer of this threat.
3D Secure 2.0: The Liability Shift You Need
3D Secure 2.0 (3DS2, also known as Verified by Visa, Mastercard Identity Check, and AmEx SafeKey) is the industry standard for authenticating cardholders in CNP transactions. When a cardholder completes 3DS2 authentication and the transaction is later disputed as unauthorized, the liability shifts from you to the issuing bank. You keep the money. This makes 3DS2 the single most important tool for any e-commerce merchant or virtual terminal user accepting card-not-present payments.
Unlike the original 3DS (which added friction with mandatory redirects and password challenges), 3DS2 performs a risk assessment invisibly in the background using device data, behavioral signals, and transaction history. Low-risk transactions, the vast majority, authenticate frictionlessly. Only high-risk transactions trigger a challenge (OTP, biometric, etc.).
The liability shift is the key benefit
With 3DS2-authenticated transactions, if a cardholder disputes the charge as unauthorized, the issuing bank absorbs the loss, not you. This directly reduces both your revenue loss and your chargeback ratio, protecting your merchant account status.
AI-Based Transaction Scoring
In 2026, the fraud arms race is squarely between AI-powered fraud tools and AI-powered fraud attacks. Fraudsters now use generative AI to create synthetic identities, deepfake voice and video for account takeover, and automated card testing at unprecedented scale. Roots' real-time fraud scoring evaluates hundreds of signals in the milliseconds between transaction submission and authorization response, including behavioral signals specifically calibrated to detect AI-generated inputs:
- Device fingerprint, is this a known device, or a new/suspicious one?
- Behavioral biometrics, typing speed, mouse movement, form completion patterns
- IP geolocation, does the IP match the billing address? Is it a VPN or Tor exit node?
- BIN (Bank Identification Number) analysis, is the card issued in the expected country?
- Velocity checks, how many transactions has this card/device attempted in the past hour?
- Order anomalies, unusually high value, rush shipping to freight forwarder, mismatched names
- Network consortium data, has this card or device been flagged by other merchants?
Velocity Rules and Block Lists
Velocity rules protect against enumeration attacks, where fraudsters test stolen card numbers with small transactions to identify live cards before making large purchases. Common controls include maximum attempts per card number, per IP address, and per device per hour. These are configurable through your fraud settings and can be tuned to your transaction patterns without generating high false-positive rates.
AVS and CVV: Basic Controls That Still Matter
Address Verification Service (AVS) checks the numeric billing address and ZIP code against the card issuer's records. CVV checks the 3–4 digit code on the card. Neither is a guarantee of legitimacy, both data points can be stolen, but they are meaningful signals in your risk scoring model and critical evidence in a chargeback representment case.
Balancing Fraud Prevention With Conversion
Overly aggressive fraud rules create false positives, legitimate customers who can't complete their purchase. A 1% false positive rate on $1M in annual e-commerce revenue means $10,000 in lost sales. The goal is a fraud stack that catches bad transactions without blocking good ones. Roots' fraud tools include a manual review queue for flagged transactions, giving your team the ability to verify borderline orders without automatically declining them. Most merchants achieve a false positive rate below 0.3%.
For a complete picture of how fraud prevention, alert networks, and chargeback management work together, see Roots' full risk mitigation suite, all four layers are available to every Roots merchant account holder.
Ready to put this into action?
Talk to a Roots Payment Solutions Expert
Free consultations, no obligation. Our experts apply exactly the concepts in this article to your specific account.