What should a bank look for in AI fraud detection software?
AI fraud-detection software scores events as they happen, such as a card payment, a login, a new-account application, a wire or a phone call, and returns a risk score or a decision within milliseconds. Banks buy it as a hosted platform priced by volume and licence, connect it to their payment and digital channels, and tune rules and models on their own data. The questions that separate vendors are latency, the data network behind the models, scam coverage, explainability for analysts, and how the vendor supports model validation.
Observations from how these products are sold and regulated, not advice.
01
Coverage of the fraud types you actually lose money to
Card-not-present, account takeover, new-account and synthetic identity, authorised push-payment and scam, check and wire fraud need different signals. A vendor strong on card scoring may have little for scams where the customer is tricked into sending money.
02
Latency and where the model runs
Payment authorisation leaves tens of milliseconds for a decision. Ask for the vendor's measured response time at your volume and whether scoring runs in its cloud, in yours or on premises.
03
The data behind the model
Consortium or network data (device, behavioural or cross-bank signals) can catch fraud a single bank's history cannot. Ask what data leaves the bank, how it is pooled and what rights the vendor keeps.
04
False positives, not only detection
Declined good customers and analyst workload are the cost of a fraud model. Vendor-reported detection gains are usually measured against a customer's prior system; ask for the false-positive ratio and the baseline behind any percentage.
05
Explainability and case management
Analysts need the reasons behind a score to work an alert, and model-risk teams need documentation to validate it. Check what the vendor provides for both, and whether you can export decisions and features.
06
Pricing that tracks your volume
Most fraud platforms are sold on annual licences tied to transaction or account volume, with implementation fees; published price lists are rare. Model the cost at next year's volume, not this year's.
Which rules apply to a bank using AI fraud detection software?
| Rule | Why it applies | Documents |
|---|---|---|
| Model risk management (SR 26-2 / OCC 2026-13 / FIL-15-2026) | Fraud models are models under the April 2026 interagency guidance; a bank above $30 billion in assets is expected to inventory, validate and monitor them, including vendor models. Generative and agentic AI are outside its scope. | SR 26-2 OCC Bulletin 2026-13 FDIC FIL-15-2026 |
| Third-party risk management | Buying a fraud platform is a third-party relationship: due diligence, contract terms, ongoing monitoring and exit planning under the 2023 interagency guidance. | OCC Bulletin 2023-17 SR 23-4 FDIC FIL-29-2023 |
| Deepfake and synthetic-identity fraud | FinCEN's 2024 alert lists red flags for generative-AI identity documents, images and audio used to open accounts, and asks for a specific SAR key term; identity and voice vendors are bought against exactly these threats. | FIN-2024-Alert004 (Deepfake Media) |
| Adverse action where fraud models decide credit | When a fraud or identity score contributes to declining a credit application, the ECOA/Regulation B and FCRA notice duties apply to that decision. | ECOA / Regulation B adverse action (15 U.S.C. 1691(d); 12 CFR 1002.9) FCRA adverse action and credit-score disclosures (15 U.S.C. 1681m, 1681g(f)) |
| EU AI Act fraud carve-out | For banks operating in the EU, AI used solely to detect financial fraud is excluded from the Act's high-risk credit-scoring category, unlike credit models. | Regulation (EU) 2024/1689 |
Documents link to the AI Regulation Tracker. By use case: Fraud detection · AML / KYC · Model risk management · Third-party & vendor AI.
How do the AI fraud detection vendors compare?
| Vendor | What it does | Deployment | Pricing model | Published price | US bank customers* | Founded · ownership |
|---|---|---|---|---|---|---|
| Feedzai | RiskOps platform for bank fraud, scam and AML prevention: transaction scoring, behavioral biometrics and network intelligence. | SaaS (cloud-based); Feedzai IQ Score is also delivered via a single API and listed on AWS Marketplace | Sales-led enterprise licensing, scoped per institution and use case (fraud, AML, onboarding, scam prevention); Feedzai IQ Score is also sold as an API-metered add-on listed on AWS Marketplace | Not published | 2 | 2011 · Private |
| Featurespace | Real-time behavioral fraud and AML scoring for banks and processors (The Featurespace Platform, formerly ARIC Risk Hub); a Visa unit. | Hosted cloud or on-premises; Visa A2A Protect is delivered through a single API | Sales-led enterprise licensing, sold as a Visa solution; A2A Protect is offered to financial institutions via a single API | Not published | 1 | 2008 · Acquired by Visa |
| BioCatch | Behavioral-biometric and device intelligence that scores digital banking sessions for account takeover, scams, mules and application fraud. | SaaS; collected through BioCatch's Align SDK in the bank's web and mobile apps, and available through Q2, Alkami and Alloy integrations | Sales-led enterprise licensing per institution, plus distribution through digital-banking and onboarding platform partners (Q2, Alkami, Alloy); no figures disclosed. | Not published | 3 | 2011 · Private |
| Sardine | Fraud and AML platform: device and behavioral signals, transaction monitoring, sanctions screening and AI agents for alert review. | SaaS, integrated through APIs and a client-side device and behavior SDK | Sales-led enterprise contracts; Sardine describes usage-based pricing for its custom agents | Not published | 0 | 2020 · Private |
| Socure | Identity verification, fraud scoring and watchlist screening for account opening and login, run on its RiskOS decisioning platform. | API (cloud); self-serve Socure Launch or configurable Socure Enterprise on the RiskOS platform (Direct API, hosted UI, SDKs) | Per-evaluation self-serve pricing (Socure Launch) with $1,000 of free monthly credits; usage-based enterprise contracts with volume discounts (Socure Enterprise / RiskOS) | $1.00 per evaluation ($1,000 of free monthly credits) (Socure Launch (self-serve bundles) — KYC + Fraud + Watchlist, DocV Step up) | 4 | 2012 · Private |
| Alloy | Identity and fraud prevention platform: onboarding KYC/KYB, fraud and login risk scoring, AML monitoring and data orchestration. | SaaS (single API, SDK and no-code dashboard; hosted on AWS) | Sales-led enterprise platform contracts; bought through a demo and quote process | Not published | 3 | 2015 · Private |
| Pindrop | Phone-channel fraud scoring, passive caller authentication and voice deepfake detection for bank contact centers. | SaaS (cloud), integrated with contact-center platforms such as NiCE CXone and Zoom Contact Center; also sold through AWS Marketplace and FICO Marketplace | 12-month contract licences per product sized by call volume (up to 1,000,000 calls per unit) or by meeting users, with per-unit charges for additional usage; custom and private offers through sales | $250,000 per 12-month contract, licence for up to 1,000,000 calls; additional usage $0.30 per unit (Pindrop Anti-Fraud and Authentication Platform, Pindrop's own AWS Marketplace listing — Anti Fraud) | 2 | 2011 · Private and venture-backed: a $75 million Series C led by Google Capital |
| NICE Actimize | Fraud, AML, sanctions screening, KYC, case management and trade surveillance software for banks, on X-Sight and Xceed platforms. | SaaS (X-Sight and Xceed cloud platforms) or on-premises under term licences | Sales-led enterprise contracts: cloud subscriptions with transaction- or usage-based fees, or on-premises term licences, plus professional services | Not published | 4 | 1999 · Business unit of NICE Ltd |
| Hawk | AML transaction monitoring, sanctions and customer screening, fraud prevention and AI investigation agents for banks and payment firms. | SaaS (multi-tenant, multi-cloud) or private cloud | Sales-led enterprise SaaS contracts; modules (AML, screening, fraud) sold separately or together under one contract | Not published | 0 | 2018 · Private |
| Unit21 | Fraud and AML platform: transaction monitoring, screening, case management and SAR filing, with AI agents that run alert investigations. | SaaS (cloud-based), accessed through Unit21's API and dashboard | Sales-led annual plans for Fraud or AML, plus implementation and support options (Standard or Gold support, Solutions Architect add-on) | Not published | 1 | 2018 · Private and venture-backed |
| Zest AI | AI credit underwriting for banks and credit unions, plus fraud detection (Zest Protect) and gen-AI lending intelligence (LuLu). | SaaS | Sales-led, per-lender enterprise contracts; not published. | Not published | 2 | 2009 · Private |
| Provenir | Cloud decisioning platform for credit risk, fraud and identity, with a data marketplace, case management and AI models. | SaaS (cloud-native; API-first, real-time and batch) | Sales-led enterprise software licence (SaaS); not published. | Not published | 1 | 2004 · Private |
| FICO | FICO Platform decisioning and analytics software for origination, account management and fraud (Falcon), plus the FICO Score. | SaaS or on-premises | Software: multi-year subscriptions priced on usage (accounts, transactions or decision use cases), often with contracted minimums; not published. FICO Score for mortgages: published per-score and funded-loan fees | Not published (prices shown for FICO Score for mortgage, Mortgage Direct License Program (not the FICO Platform) on the vendor page) | 2 | 1956 · NYSE: FICO |
* Banks among the 100 largest US banks on this site publicly named as customers, each sourced on the vendor's page. Prices are shown only where the vendor publishes them for the AI product itself; a price for another product (a credit score, a notary plan, contact-centre seats) is listed on the vendor's page, not here. More vendors in this category are being researched.
Private · founded 2011
Feedzai
Sells banks a single AI-native RiskOps platform, rather than point tools, for fraud, scam and AML prevention, extending in 2025-2026 into GenAI scam agents, federated network intelligence and a tabular foundation model for risk decisions.
NYSE: V · founded 2008
Featurespace
Sells banks, card processors and merchant acquirers a real-time machine-learning platform that profiles each customer's normal behavior to score payments, cards, checks, applications and AML activity; since the 2024 acquisition it is sold as a Visa solution and its technology is being folded into Visa products such as A2A Protect.
Banks among the 100: U.S. Bank
Private · founded 2011
BioCatch
Sells banks behavioral-biometric and device intelligence that runs inside their digital banking sessions to flag account takeover, scams, money-mule accounts and account-opening fraud, licensed directly and through digital-banking and onboarding platforms; being acquired by Visa.
Banks among the 100: Wells Fargo, BofA, Citi
Private · founded 2020
Sardine
Sells banks, fintechs and merchants one modular platform for fraud prevention and AML compliance, built on device and behavioral signals and a cross-industry data consortium (Sonar), and since 2025 adds AI agents that triage alerts and draft investigation work under human review.
Banks among the 100: none named publicly yet
Private · founded 2012
Socure
Sells banks, fintechs and government agencies API-based identity verification, identity-fraud scoring and sanctions screening, now packaged in the RiskOS decisioning platform and charged per evaluation or on usage-based enterprise contracts.
Banks among the 100: Capital One, Citi, SoFi, Atlantic Union
Private · founded 2015
Alloy
Sells banks, credit unions and fintechs a SaaS identity and fraud prevention platform that orchestrates more than 270 third-party data solutions behind one API, adding its own machine-learning models and, since 2026, an agentic AI Assistant for risk and compliance reviews.
Banks among the 100: Ally, Banc of California, Eastern Bank
Private · founded 2011
Pindrop
Sells banks, insurers and healthcare payers cloud software that scores every phone call for fraud risk, authenticates callers passively from voice, device and network signals, and detects synthetic voices and AI agents, sold directly, through contact-center platforms (NiCE, Zoom), AWS Marketplace and FICO Marketplace.
NASDAQ: NICE · founded 1999
NICE Actimize
Sells banks fraud, anti-money-laundering, sanctions-screening, KYC/CDD, case-management and markets-compliance applications on two platforms, X-Sight for large institutions and Xceed for the mid-market, as cloud subscriptions or on-premises term licences, and since 2025 adds agentic-AI investigation features and a cross-institution risk network.
Banks among the 100: KeyBank, Associated, Fifth Third, Valley
Private · founded 2018
Hawk
Sells banks and payment firms a modular, cloud-based platform for AML transaction monitoring, screening and fraud prevention that can replace a rules engine or sit on top of one as an AI overlay, adding in 2026 AI agents that investigate AML alerts and review screening hits.
Banks among the 100: none named publicly yet
Private · founded 2018
Unit21
Sells fintechs, sponsor banks, community banks and credit unions one cloud platform for fraud and AML (detection rules, screening, case management and regulatory filing), and since 2025 has rebuilt it around AI agents that investigate alerts and draft SAR narratives for human approval.
Banks among the 100: Sallie Mae
Private · founded 2009
Zest AI
Sells AI-automated underwriting models, priced and built per lender, to banks, credit unions and specialty lenders, and cross-sells fraud detection (Zest Protect) and a generative-AI lending-intelligence assistant (LuLu) into the same customer base.
Banks among the 100: Zions, First Hawaiian
Private · founded 2004
Provenir
Sells banks, credit unions, fintechs and non-bank lenders a configurable decisioning platform, licensed as cloud software, on which the lender runs its own credit, fraud and identity strategies and models, with third-party data supplied through Provenir's marketplace.
Banks among the 100: SoFi
NYSE: FICO · founded 1956
FICO
Sells banks two things: the FICO Score, licensed per score mostly through the credit bureaus, and decisioning software (FICO Platform, Falcon fraud, originations and account-management solutions) on multi-year subscriptions priced by accounts, transactions or use cases.
How much does AI fraud-detection software cost a bank?
Most vendors do not publish prices. Enterprise fraud platforms are sold on annual licences scaled to transaction, account or API-call volume, plus implementation. Where a price is public it is on the vendor's own pricing page and is listed in the comparison table above with its date; otherwise the table says 'Not published'.
Is a fraud model subject to model risk management rules?
Yes. The April 2026 interagency guidance (Fed SR 26-2, OCC Bulletin 2026-13, FDIC FIL-15-2026) covers quantitative models used in decisions, including vendor fraud models; it is expected to matter most for banks above $30 billion in assets and leaves generative and agentic AI outside its scope.
What is the difference between fraud detection and AML software?
Fraud detection protects the bank and its customers from losses in real time, usually before a payment completes. AML software monitors activity after the fact for money laundering and sanctions exposure, and feeds suspicious activity reports. Several vendors now sell both on one platform, which the industry calls FRAML.
Which AI fraud detection vendors are compared here?
Feedzai (RiskOps platform for bank fraud, scam and AML prevention: transaction scoring, behavioral biometrics and network intelligence); Featurespace (Real-time behavioral fraud and AML scoring for banks and processors (The Featurespace Platform, formerly ARIC Risk Hub); a Visa unit); BioCatch (Behavioral-biometric and device intelligence that scores digital banking sessions for account takeover, scams, mules and application fraud); Sardine (Fraud and AML platform: device and behavioral signals, transaction monitoring, sanctions screening and AI agents for alert review); Socure (Identity verification, fraud scoring and watchlist screening for account opening and login, run on its RiskOS decisioning platform); Alloy (Identity and fraud prevention platform: onboarding KYC/KYB, fraud and login risk scoring, AML monitoring and data orchestration); Pindrop (Phone-channel fraud scoring, passive caller authentication and voice deepfake detection for bank contact centers); NICE Actimize (Fraud, AML, sanctions screening, KYC, case management and trade surveillance software for banks, on X-Sight and Xceed platforms); Hawk (AML transaction monitoring, sanctions and customer screening, fraud prevention and AI investigation agents for banks and payment firms); Unit21 (Fraud and AML platform: transaction monitoring, screening, case management and SAR filing, with AI agents that run alert investigations); Zest AI (AI credit underwriting for banks and credit unions, plus fraud detection (Zest Protect) and gen-AI lending intelligence (LuLu)); Provenir (Cloud decisioning platform for credit risk, fraud and identity, with a data marketplace, case management and AI models); FICO (FICO Platform decisioning and analytics software for origination, account management and fraud (Falcon), plus the FICO Score). Each has its own page with products, pricing, ownership and the banks among the 100 largest US banks that are named as customers. More vendors in this category are being added as their pages are researched and sourced.
When a bank picks an AI fraud detection vendor, you'll read it here first.
when a bank picks an AI vendor, or a vendor ships for banks, it is in the next morning's brief · six sourced stories · 7 am ET · free
plus every tracker, bank and agent page update, the morning after · leave any morning