Ecuador SRI — Sistema Falcon: AI-Powered Tax Risk Scoring

The Servicio de Rentas Internas (SRI), Ecuador's national tax authority, has operated Falcon since 2020, a machine-learning risk-scoring platform that cross-references more than 400 internal and external data types — including property registries, financial transactions, social security records and commercial databases — to assign each taxpayer a score from 0 to 1,000. Higher scores trigger prioritised audit attention and targeted compliance outreach. The initiative was publicly presented on 2 October 2023, when President Guillermo Lasso signed an executive decree declaring tax and customs modernisation a national priority.

Falcon aims to close Ecuador's tax gap by using risk-based targeting to identify non-compliant and unregistered taxpayers rather than relying on broad, undifferentiated audits. A parallel SRI study, developed with International Monetary Fund advisory support, estimated the country's annual personal income-tax gap to establish the fiscal baseline against which Falcon's compliance impact will be measured over time.

Falcon's scoring engine draws on more than 400 data sources to flag taxpayers for audit and compliance outreach, and its rollout is being scaled as part of the broader Orión modernisation programme, financed by a USD 60 million Inter-American Development Bank loan, which also covers AI-driven audit selection, cybersecurity upgrades and hardware renewal through 2027. An independent TADAT (Tax Administration Diagnostic Assessment Tool) review assessed the SRI's risk-management practices, and separate academic studies applied Random Forest and spatial-clustering methods to SRI's taxpayer records to validate the algorithmic approach.

By October 2023, Falcon had identified approximately 60,000 individuals holding property or other income-generating assets who had never registered with the tax system. The SRI/IMF study estimated Ecuador's annual personal income-tax gap at USD 1.121 billion, and the TADAT review awarded the SRI an 'A' grade for risk identification and a 'B' for risk classification and management. Independent academic analyses applying Random Forest and spatial-clustering methods to 2.5 million taxpayer records achieved 92–93% accuracy in predicting non-compliance risk.

Academic reviewers note that no revenue figure has been independently attributed to Falcon itself, and they flag concerns over limited algorithmic transparency, due-process guarantees, and the handling of sensitive taxpayer data under Ecuador's 2021 Personal Data Protection Law. Full AI integration under the Orión programme is targeted for completion by 2027, meaning the system's mature impact is still to be established.