AI Risk Screening for Customs — Tunisia's National Selectivity System Targets Fraud

On 16 May 2026, Tunisia's Central Committee for Judicial Risk Action, under the Direction Générale des Douanes (Customs Authority), approved integrating a machine-learning module into the national 'Système de Sélectivité' used to screen import and export declarations.

Score incoming shipments by analysing the nature of goods, country of origin, declared value, import history and the profile of the economic operator, to flag high-risk consignments for physical inspection while letting low-risk, compliant trade clear faster.

The module was integrated directly into the existing national selectivity system covering all customs points, alongside two related 2026 initiatives: 'Sinda 2', a next-generation platform connecting government agencies on customs data, and a joint project with South Korea to digitise customs procedures.

Tunisian customs reported nearly 4,000 enforcement cases and over 51 million dinars (about $17.6 million) in seized goods for Q1 2026 — but that reporting period ends before the AI module's May 2026 approval, so these figures describe general enforcement performance rather than a measured before/after effect of the new algorithm. No public accuracy, false-positive, or inspection-rate figures for the AI module itself have been published.

The deployment reflects a broader institutional push into digitised, data-driven customs enforcement, but the AI module's own effect on detection accuracy or trade flow has not yet been independently measured or disclosed.