Essay

Why fraud models should only propose. By Amir Shachar.

Read the essay
Skip to content
Riskernel

Blog · Fraud infrastructure

Fraud infrastructure notes for teams that need answers before they need a vendor.

Practical evaluation: how the main vendors compare, what to test before you buy, and where losses hide in payouts, onboarding and AI agents.

Alternatives

  1. Sardine Alternatives for Fintechs: 2026 Guide

    When Sardine is the right choice, when it is not, and how Unit21, SEON, Alloy, Sift, an in-house build and Riskernel compare for fintechs in 2026.

  2. Unit21 Alternatives for Fintechs: 2026 Guide

    When Unit21 is the right choice, when it is not, and how Sardine, SEON, Feedzai, an in-house build and Riskernel compare for fintechs in 2026.

  3. NICE Actimize Alternatives for Fintechs: 2026 Comparison

    Comparing NICE Actimize alternatives for fintechs in 2026: SEON, Sardine, Unit21, Feedzai and Riskernel on pricing model, latency, explainability and speed.

  4. SEON Alternatives for Fintechs: 2026 Guide

    When SEON is the right choice, when it is not, and how Sardine, Unit21, Alloy, Sift, an in-house build and Riskernel compare for fintechs in 2026.

  5. Sift Alternatives for Fintechs: 2026 Guide

    When Sift is the right choice, when it is not, and how Sardine, SEON, Unit21, Alloy, an in-house build and Riskernel compare for fintechs in 2026.

  6. Alloy Alternatives for Fintechs: 2026 Guide

    When Alloy is the right choice, when it is not, and how Sardine, Unit21, SEON, Feedzai, an in-house build and Riskernel compare for fintechs in 2026.

  7. Feedzai Alternatives for Fintechs: 2026 Guide

    When Feedzai is the right choice, when it is not, and how Sardine, Unit21, SEON, Alloy, NICE Actimize and Riskernel compare for fintechs in 2026.

  8. Build Fraud Detection In-House? When It Wins and When It Does Not

    When building fraud detection in-house is the right call, what it takes in people and time, and where outside help pays, with every source labeled.

  9. Fraud Consultancy Alternatives: Boutiques, Platforms and a Lab

    Fraud and fincrime consultancies and boutiques compared with platforms and with a lab like Riskernel, with where each is the better choice and sources.

Field notes

  1. Vendor Bank-Detail Changes: The Callback Control That Stops Payment Redirection, and How to Prove It Works

    Why payment-redirect fraud succeeds, how the callback-and-second-approval control fails in practice, and how to test and evidence it.

  2. Free Credits Are Compute. Who Is Spending Yours?

    How free-credit farming, multi-accounting, key resale and metering gaps drain usage-priced AI and API platforms, with a self-check and how the audit works.

  3. Your Scam Warnings Pass Review. Do They Stop a Customer Mid-Scam?

    How to test whether APP scam warnings hold up against a coached customer, with a self-check your fraud team can run, and how the Scam Warning Lab works.

  4. Can Your AI Agent Be Talked Into Moving Money? Testing Chat and Voice Agents Against Scam Scripts

    How fraudsters social-engineer customer-facing AI chat and voice agents, where those agents fail, and a self-check fraud teams can run.

  5. Deepfake Scam Ads Get Paid Somewhere. Is It Through Your Merchants?

    Why scam-ad funnels settle through merchant accounts that looked clean at onboarding, how acquirers can check their own portfolio, and how a trace works.

  6. Verify Before You Advance: Double-Pledged Invoices and Payment Redirects in Receivables Finance

    Why factors and asset-based lenders should confirm invoices with the debtor's real payment approver before funding, plus a seven-question self-check.

  7. Your LLM May Have Already Seen the Test: Validating AI Models With Held-Out Exams

    Public benchmarks may sit in an LLM's training data. How model-risk and fraud teams can build held-out, code-checked tests that support effective challenge.

  8. What Happens to Your Fraud Defence If Your Vendor Disappears Tomorrow?

    Most fintechs depend on one outside fraud model and have never tested life without it. Why regulators expect an exit plan, and seven questions to answer.

  9. The Fake CEO on the Video Call: Would Your Finance Team Pay?

    How deepfake executive impersonation beats payment controls, what the Arup case shows, and a self-check for finance, treasury and payroll.

  10. Field Exams Sample a Handful of Invoices. The Fraud Is in the Rest of the Book.

    Why sampled field exams miss double-pledged and fictitious receivables, and eight whole-book tests lenders can run with data they already have.

  11. How Do You Know Your AI Fraud Investigator Is Right? Testing Agents Against a Known Answer Key

    AI agents can write convincing fraud and AML dispositions. To know if they are right, you need cases with known truth. A guide for fraud and AML leaders.

  12. Your AI Feature Is Stuck in Bank Vendor Review. Written Answers Won't Unstick It.

    What bank and insurer reviewers check in an AI feature, why written assurances stall deals, and a seven-step self-test for prompt injection and data leakage.

  13. Who Decides When Your AI Agent Goes Wrong? Running an Incident Tabletop Before the Regulator Asks

    AI agent incidents do not look like outages or breaches. How to rehearse decision rights, with a self-check any bank or fintech risk team can run.

  14. What Is the Worst Your AI Agent Could Do? Putting a Money Number on It

    Most AI agent reviews test the model. Risk committees need the worst-case loss if an agent is hijacked. How to estimate it from tool scopes and limits.

  15. Shadow Testing a Fraud Vendor Before You Touch Production

    A practical guide to shadow testing fraud vendors: what to compare, how long to run it, and which red flags show up before production.

  16. First-Time Payees, Payouts, and Why Clean Transactions Still Turn Into Fraud Losses

    Why many fraud losses do not look suspicious at the transaction level, and what teams miss when they score events without scoring the setup around them.

  17. Handling Extreme Class Imbalance in Fraud Detection

    Why fraud datasets break standard ML assumptions, and what teams should measure instead of trusting impressive offline metrics.

  18. SHAP Explainability for Fraud Ops: What Analysts Actually Need

    A practical guide to SHAP explainability in fraud operations: what analysts need to see, where feature importance falls short, and how reasons speed up review.

  19. Real-Time Fraud Scoring Latency: What a 47 ms Median Actually Means

    A practical guide to fraud scoring latency: what the median, 95th and 99th percentile mean in production, why tail latency matters, and how to test vendors.

  20. Fraud Detection API: What to Look For in 2026

    A practical guide to evaluating fraud detection APIs in 2026, including latency, explainability, shadow testing, integration risk, and analyst usability.

Signal Log

Fraud engineering notes, biweekly and numbered. See the newsletter.

Start with the attack tree of your own launch.

A Launch Fraud Review is a private attack tree of your new flow. Free, in writing. Runs beside your current system. Nothing to rip out.