DOZORRO — Ukraine's AI Risk-Flagging System for Public Procurement

Prozorro is Ukraine's mandatory electronic public-procurement system; DOZORRO is the civic-monitoring layer built on top of it by a coalition led by Transparency International Ukraine, in cooperation with the Ministry of Economy.

Use machine learning, trained on how procurement experts assessed tenders, to flag corruption-risk patterns that a fixed checklist of indicators could be learned and circumvented by officials.

From July 2018, developers trained the model on roughly 3,500 tenders independently assessed by 20 procurement experts. The tool went into beta in November 2018 and full operation in Q1 2019. High-risk tenders identified by the model are routed to civil-society monitors, who decide whether to escalate to auditors or law enforcement.

In beta testing the model flagged 26% more tenders with unfounded winner selection, 37% more with groundless disqualification, and 298% more apparent collusion cases than the State Audit Service's separate system of 35 fixed risk indicators. In July 2025 alone, monitors analysed 165 procedures worth over UAH 10 billion and found violations in 89 of them; across H1 2025 the coalition's interventions helped prevent an estimated UAH 133 million in inefficient spending.

The underlying model's exact features and false-positive rate are not publicly documented, limiting independent verification of the beta-test percentages beyond the developers' own account.