Artificial Intelligence and Digital Public Oversight in Preventing Corruption: A Governance Framework for Uzbekistan

Uzbekistan is simultaneously expanding digital government, open-data obligations, electronic procurement, corruption-risk assessment and artificial-intelligence capacity. These reforms create an opportunity to move from periodic, complaint-driven control towards continuous prevention. They also create a new governance problem: an opaque risk-scoring system can reproduce poor data, conceal institutional discretion, stigmatise innocent persons and relocate corruption from frontline officials to model design, data access and vendor procurement. This article develops TRACE-UZ, a country-adapted framework for using artificial intelligence in public oversight without delegating legal judgment to an algorithm. The framework was constructed through doctrinal and policy analysis of Uzbekistan’s anti-corruption, openness, procurement, audit and AI instruments, read alongside the OECD’s fifth-round monitoring report, the United Nations Convention against Corruption, UNESCO’s Recommendation on the Ethics of Artificial Intelligence and comparative procurement-analytics literature. TRACE-UZ contains five mutually dependent layers: Transparent data foundations; Risk analytics; Accountable human review; Civic contestability; and Evaluation and institutional learning. It connects machine-readable procurement and budget data with rules, anomaly detection, network analysis and natural-language processing, while requiring reason codes, human verification, protected reporting, appeal channels, public performance reporting and independent audit. The article proposes a phased implementation roadmap, an institutional allocation of responsibilities and a measurement architecture that separates operational accuracy from public-value outcomes. Its central claim is that artificial intelligence can strengthen integrity only when embedded in a public oversight loop: data produce risk signals; authorised humans investigate; affected persons can contest; civil society can scrutinise system performance; and verified outcomes improve both institutions and models. AI should triage risk, not determine guilt.