On 1 February 2023, Armenia's Corruption Prevention Commission (CPC) — an independent five-member body established in November 2019 — launched the Electronic Platform for Declarations (EPD) of assets, incomes, expenditures and interests, funded by the World Bank and built by Synergy on its IDM platform. The system auto-populates declaration fields from Armenia's State Revenue Committee, cadastral, civil and police registries, and had, per the vendor's own case-study figures, registered 31,579 users, created 41,309 official profiles, and processed 12,976 annual declarations for 2022 alone.
CPC Chair Haykuhi Harutyunyan has said artificial intelligence is enhancing the Commission's ability to scrutinize the asset and interest declarations filed by politicians and public servants, making it easier to detect potential cases of fraud and malfeasance, and that "data-driven technologies can be a force for accountable governance" provided institutions build trust with stakeholders across government, civil society and the public.
The Commission made a deliberate transparency trade-off: raw declaration data is publicly viewable by anyone, but the public cannot see whether the AI tools have flagged a specific declaration as suspicious — intended to protect the integrity of ongoing reviews but also limiting independent scrutiny of the AI's own flagging decisions. The Council of Europe has provided technical support to CPC staff on verifying and analysing declarations as part of this effort.
Public reporting to date describes the AI capability qualitatively — as an enhancement to analyst scrutiny — without publishing quantified detection rates, false-positive rates, or case outcomes attributable specifically to the AI layer, and the CPC has also faced practical hurdles engaging local IT vendors to build out platform features due to public-procurement integrity concerns.
Read the full analysis: https://eurasianet.org/perspectives-artificial-intelligence-boosts-anti-corruption-efforts-in-armenia
Where this practice's information was retrieved from, and when.