Publican AI — Ghana’s machine-learning customs enforcement at Tema Port

The Ghana Revenue Authority (GRA) deployed Publican AI at Tema Port — Ghana’s primary gateway for imports — from 1 February 2026. The system uses machine-learning models to analyse import declarations, benchmark stated values against global trade price databases, and flag anomalies for human review before shipment clearance. GRA describes it as decision-support technology: Publican AI assigns a risk score to each declaration, with approximately 25% flagged for additional scrutiny, reducing manual processing time from around two hours to five minutes per declaration.

In its first weeks of operation, GRA data showed monthly customs revenue rising from GH₵2.4 billion to GH₵3.6 billion — a near-50% increase attributed largely to improved import valuation. By April 2026, GRA claimed the system was generating approximately USD 3 million per day in additional revenue, and cited historical estimates of GH₵11 billion in annual revenue lost to valuation fraud and under-invoicing over 2020–2024.

However, independent analysts publicly questioned whether the revenue surge was causally attributable to the AI system rather than other enforcement factors, calling the evidence “too early to verify” and requesting a formal independent audit. Operational concerns have also emerged: traders at Tema Port report they cannot understand how customs values are determined, disputes are escalated to a secretariat that meets only twice weekly, and no published algorithmic transparency exists.

In response to stakeholder concerns, the Ministry of Finance convened a multi-party technical committee in April 2026 to oversee the system. No independent audit had been completed as of June 2026. This case is included as an early-stage deployment with promising but contested results and documented governance gaps — a pattern the C-NAPSE rubric scores conservatively.

Read the full analysis: https://gra.gov.gh/