Turkey VDK KURGAN — AI real-time anomaly detection in e-invoices for tax fraud

Turkey's Tax Audit Board (Vergi Denetim Kurulu, VDK) launched the KURGAN anomaly-detection system on 1 October 2025 as the centrepiece of its Fake Document Combat Strategy (Sahte Belgeyle Mücadele Stratejisi). Applying machine learning across 13 evaluation criteria, KURGAN cross-references every e-invoice, e-ledger, and banking transaction in near real time, flagging purchase-sale mismatches, warehouse-capacity discrepancies, and patterns indicative of carousel or missing-trader VAT fraud.

Rather than issuing automatic penalties, KURGAN sends early-warning letters ("KURGAN letters") to flagged taxpayers, giving them the opportunity to correct declarations before formal enforcement begins. This advisory-first philosophy is designed to raise voluntary compliance alongside detection rates.

Turkey's mandatory e-invoicing infrastructure — spanning millions of VAT-registered businesses and billions of documents annually — provides the data substrate that makes KURGAN's nationwide, real-time cross-referencing feasible.

In 2025, VDK's AI-assisted compliance programme — of which KURGAN is the primary real-time detection component — identified approximately 748 billion Turkish lira in undeclared taxable income and assessed 294 billion TL in tax penalties, while 86.8 billion TL in additional tax base was voluntarily declared through the invite-to-explain process. Because KURGAN launched on 1 October 2025, full-year figures reflect a blend of legacy and new-system detection, so outcome attribution to KURGAN specifically remains partial.

Read the full analysis: https://www.duygusener.com/en/post/kurgan-system-and-turkeys-ai-supported-public-auditing