# Catch the account, then explain why: an AI pipeline for money mules

> Researchers pair a fraud-detection model with SHAP attribution and an open-weight LLM that writes the case in plain English — and report catching far more mule accounts in live production.

_Source: arXiv preprint (peer review pending) — authors' own production results · 2026-07-22 · 6 min read · Verified against primary sources_

Canonical: https://iyu.app/e/explainable-money-mule-pipeline

## Full explainer

> **⚑ Caveat:** This is a **non-peer-reviewed preprint**, and the headline numbers (89% yield, the 61% baseline, the 60% uplift) are the **authors' own results from a single live deployment** — promising, but not independently reproduced. Treat the metrics as self-reported.


### The problem — What's a money mule?

A **money mule** is a bank account used to move illicit funds — the middle link that launders stolen money into the wider financial system. Sometimes the account holder is a knowing accomplice; often they've been recruited or duped. Mule accounts are a key enabler of fraud, and catching them at scale is hard because the evidence is a messy blend of transaction patterns, account details, network structure (who pays whom), and timing.


### The idea — Detect → Attribute → Narrate


---

_This is a members-only explainer; the excerpt above is the free preview. Full text: https://iyu.app/e/explainable-money-mule-pipeline_


## Primary sources

- [Zhang et al., “Detection, Attribution, Narration: An End-to-End Pipeline for Explainable Money Mule Identification” (arXiv:2607.17586)](https://arxiv.org/abs/2607.17586)

---
_Published by iyu (https://iyu.app) — the day's AI news, checked against primary sources and rewritten in plain language. Free to quote with attribution and a link to the canonical URL._