Skip to content

Blog

Notes on fraud decisions.

How decisions get made, where AI fits and where it does not. Every outside claim links to its source.

How FRAPE decidesFour steps from top to bottom: an event is scored; rules and lists settle the clear cases; Decision Core advises only on uncertain events; your policy returns approve, challenge, review or decline.01Eventpayment, sign-up, login, profile change or payout02Rules and listsknown fraud and clear cases settled first03Decision Core advisesonly when the rules score is uncertain04Your policy decidesapprove, challenge, review or decline
The path every event takes. The posts below explain each step and the fraud it is built to stop.

Latest

All posts

  • Scrabble tiles spelling credit and risk on a wooden table
    Photo: Markus Winkler on Unsplash

    · 6 min read

    What a decision model is, and how it helps fraud teams

    A decision model takes typed facts and answers closed questions with probabilities. How it differs from rules, scorecards and classifiers, and why it suits the grey zone of fraud.

  • Scrabble tiles spelling the names of different languages
    Photo: Markus Winkler on Unsplash

    · 6 min read

    Token models versus decision models for fraud detection

    Generative models write text one token at a time; decision models answer closed questions with probabilities. Which fits a real-time fraud decision, and where each one helps.

  • Smartphone with a lock screen on yellow fabric
    Photo: Franck on Unsplash

    · 6 min read

    The case for challenge: step-up checks for the events you are unsure about

    Approve or decline is not the only choice. How a challenge, such as a passkey, 3-D Secure or a one-time code, settles uncertain events, what the evidence says, and how FRAPE closes the loop.

  • Laptop keyboard glowing in blue light in a dark room
    Photo: Philipp Katzenberger on Unsplash

    · 4 min read

    Fraud rings and the network graph: what one event cannot show you

    A single event can look clean while the accounts behind it share one device, card or phone. How linking identities exposes fraud rings, what the data says, and how FRAPE does it.

  • Graffiti on a wall that reads fake
    Photo: Markus Spiske on Unsplash

    · 6 min read

    Fraud at sign-up: synthetic identities, money mules and fake accounts

    Identity fraud is the most common case type in the UK National Fraud Database. What Cifas, UK Finance and the Federal Reserve say about synthetic identities and mules, and what helps at sign-up.

  • Person holding a payment card while using a laptop
    Photo: rupixen on Unsplash

    · 6 min read

    Card-not-present fraud in Europe: why the payment channel matters

    Most card payments in Europe happen in person, yet most card fraud happens online. What the EBA, the ECB and UK Finance report on remote card fraud, strong authentication and exemptions.

  • Hands typing on a laptop keyboard
    Photo: Glenn Carstens-Peters on Unsplash

    · 6 min read

    Account takeover and payment scams: why the moment of decision matters

    Criminals take over an account or talk its owner into paying them. What 2025 figures from UK Finance, Cifas, the FBI, the EBA and the ECB show, and why the decision must come first.

  • Banknotes from several countries scattered across a table
    Photo: Jason Leung on Unsplash

    · 6 min read

    Online fraud in numbers: what the latest official reports show

    The newest US, UK and European fraud figures side by side: what the FBI, the FTC, UK Finance and the EBA and ECB reported, what each one measures, and what they agree on.

Photo credits

All photos are used under the Unsplash License. Diagrams are drawn by FRAPE from the cited sources.

Get started

Start stopping fraud today.

Start free with test keys and the AI switched on, and see every decision explained on your own traffic. Or tell us what you need to protect, and we will walk you through how FRAPE would decide it. Compare plans.

$ curl https://api.frape.io/v1/score …Start freeTalk to sales