Colombia's Contraloría Deploys ML Risk Model to Flag Pre-Election Public-Contracting Spikes

On 19 May 2025, Colombia's Contraloría General de la República — the national supreme audit institution — launched a predictive analytics model, built with Universidad de Antioquia, to monitor public contracting for corruption risk, run by the Contraloría's Dirección de Información, Análisis y Reacción Inmediata (DIARI).

Flag contractors with patterns such as repeated non-compliance, falsified insurance policies or premature contract settlements, and specifically watch for contracting abuse tied to Colombia's electoral-spending restrictions (Ley de Garantías Electorales).

The model applies decision-tree and random-forest techniques to more than 13,000 linked data sources, including historical SECOP I/II contracting records from 2014-2025, fiscal and disciplinary sanction registries, and consortium-membership data. A derivative tool, the Modelo Analítico de Seguimiento Preventivo, continuously watches contracting activity during electoral periods.

In its first documented use, the Contraloría reported that in the week of 1-7 November 2025 — immediately before electoral restrictions took effect — 9.0 trillion pesos in public contracts were committed, with 6.1 trillion pesos (68% of that week's total) concentrated on a single day, a 190% spike against historical patterns, prompting a public alert. A follow-up run in mid-2026 flagged 14,414 contracts worth 4.55 trillion pesos signed in the 30 days after the presidential run-off, a 22.5% year-on-year increase, over 11,000 of them awarded without competitive bidding.

The evidence available is preventive and early-stage: no independent source confirms that contracts have been suspended, sanctions imposed, or funds recovered as a direct result of the alerts, and the Contraloría's own methodology and accuracy figures have not been independently audited.