evidoria

← Back to browse

Good practice Imported

STIR — Poland's Real-Time AI System for VAT Fraud Detection

Poland · Warsaw · See the Poland profile · See the Warsaw profile

Evidence: Observational / pre–post Top 46% 60/100 · Ask Evidence Copilot about this practice

Poland's tax authority runs ML risk-scoring on daily bank transaction data to flag VAT carousel fraud in near-real time and freeze suspect accounts — credited with narrowing the VAT Gap, though the algorithm is undisclosed and has frozen innocent businesses' accounts.

6.6 € billion
Poland's VAT Gap, 2017
1.7 € billion
Poland's VAT Gap, 2021
STIR — Poland's Real-Time AI System for VAT Fraud Detection

Details

Maturity
Established
Promoter
Krajowa Administracja Skarbowa (KAS / National Revenue Administration)
Period
2018–present
Keywords
tax compliance, financial crime, machine learning, banking data

Context

Operational since 2018, Poland's STIR system has banks report account/transaction data daily to a Central Register of Tax Data, against which machine-learning risk-scoring algorithms flag VAT carousel fraud in near-real time, letting tax authorities freeze suspected accounts for 72 hours (extendable to 3 months by court order).

Results

Poland credits STIR, alongside e-invoicing, with narrowing its VAT Gap from €6.6 billion (2017) to €1.7 billion (2021) — one of the EU's largest reductions — but the scoring algorithm is not publicly disclosed, and AlgorithmWatch and a peer-reviewed Intertax study document cases of legitimate businesses having accounts frozen based on opaque risk indicators.

Implementation
Indicative cost
Medium (€50k–€500k)
Time to results
Medium (1–3 years)
Staffing & skills
Krajowa Administracja Skarbowa (KAS), Clearing house (Izba Rozliczeniowa)

Conditions for success

  • Daily bank reporting pipeline to a Central Register of Tax Data

Common failure modes

  • Opaque scoring algorithm has frozen accounts of legitimate businesses

Commonly funded by

National / regional programmes

Indicative funding routes for practices of this type — always check each programme's current calls and eligibility rules.

Do you run this practice? Claim it — verified implementers get a public contact pathway and can propose corrections.

Data sources

Where this practice's information was retrieved from, and when.

Attachments

Similar practices you may find useful