Fingerprint launches Fraud Score, which caught more fraud in one test

Fingerprint, a company that identifies devices, launched Fraud Score. It is a tool that predicts fraud for each customer. It learns from the fraud outcomes that the customer reports, and it uses device data. The company says that teams can use it without their own models. In one test, it caught almost 50% more fraud. That result comes from one customer only.

How it works

Each customer gives Fingerprint its fraud outcomes. Fingerprint adds device signals and makes one model for each type of fraud. The types include payment fraud and account takeover. In account takeover, a criminal gets into the account of a user.

Teams can put the scores in their decision systems with the Fingerprint API. The models learn again when customers give new outcomes.

The numbers

In one test, Fraud Score caught almost 50% more fraud than the model that one customer used before. The number of good users that it stopped did not increase. Fingerprint gave no data about the test and no results for other customers.

What is not known

Flink, a customer of Fingerprint, says that use of only the visitor ID did not stop all fraudsters. No test results from Flink are given. It is not known how much fraud data a customer must have.

Insight: Fraud tools now compete on who owns the device signal

Fingerprint sells fraud scores from device data. In 2026, Sardine launched a tool. It sends the device data and the IP address of the person behind an AI agent to a merchant. Maybe each company wants the device signal that each fraud decision uses, because the owner gets control of buyers. Find out if you can move that signal to a different vendor.