evidoria

← Back to browse

Good practice Imported

CIAT e-IAD — Costa Rica’s AI anomaly detector for e-invoice fraud

Costa Rica · San José · See the Costa Rica profile · See the San José profile

Evidence: Observational / pre–post Top 25% 73/100 · Ask Evidence Copilot about this practice

Costa Rica’s tax directorate uses the open-source CIAT e-IAD ML model to detect anomalies in e-invoices. By 2026 it exposed a ~₡80 billion fraud network of ~200 taxpayers and recovered ₡8 bn in 2025. CIAT has since deployed it in Guatemala and Rio Grande do Sul.

~200
Taxpayers implicated in fraud network exposed (by April 2026)
~₡80 billion (~USD 150 million)
Value of false vouchers uncovered (by April 2026)
~₡8 billion (~USD 15 million)
Tax revenue recovered (2025)
30
DGT inspectors covering largest taxpayers
12 months
Model development period
CIAT e-IAD — Costa Rica’s AI anomaly detector for e-invoice fraud

Details

Maturity
Established
Promoter
Dirección General de Tributación, Ministerio de Hacienda
Period
2022–present
Keywords
tax compliance, e-invoicing, unsupervised machine learning, anomaly detection, CIAT

Context

Costa Rica's Dirección General de Tributación, with CIAT and Microsoft, deployed e-IAD, an open-source unsupervised machine-learning model co-developed by tax experts from seven CIAT member countries over 12 months, which analyses millions of e-invoices, tax returns and taxpayer-registry data to flag outliers and behavioural patterns consistent with organised tax evasion.

Results

By April 2026, the system helped DGT investigators uncover a fraud network of approximately 200 taxpayers involving false vouchers totalling roughly ₡80 billion (~USD 150 million); in 2025 alone, AI-assisted audits recovered approximately ₡8 billion (~USD 15 million) deducted irregularly through unsubstantiated electronic invoices. CIAT has since deployed the same solution in Guatemala and Rio Grande do Sul, Brazil.

Implementation

Implementation detail (cost, timeline, staffing, conditions for success) is not yet available for this practice.

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