European Commission supports Lithuania's STT with AI-driven fraud detection in public spending; doubles accuracy, fivefold more positives in risk tier

European Commission supports Lithuania’s STT with AI-driven fraud detection in public spending; doubles accuracy, fivefold more positives in risk tier

AI-driven fraud detection and digital transformation in law enforcement - Reforms and Investments

Supporting reforms that help unlock digital growth potential and deploy innovative solutions for citizens and businesses

AI-driven fraud detection and digital transformation in law enforcement

The European Commission provided technical support to the Special Investigation Service of Lithuania (STT) to improve its data management as well as design and implement an AI-driven methodology for detecting fraud and corruption risks in public spending. Through explainable AI, the initiative aims at improving case prioritisation, supporting investigations, and reinforce integrity, accountability, and effectiveness of national and EU funds.

This project arose within Lithuania’s Recovery and Resilience Plan (“Next Generation Lithuania”), and aligns with EU priorities on Digital Transformation and strengthening public sector governance under the Recovery and Resilience Facility. While Lithuania ranks among the EU’s most digitally advanced administrations, challenges remain in leveraging digitalisation to enhance transparency and anti-corruption controls. Key issues included fragmented public sector data, limited interoperability, and insufficient use of advanced analytics for fraud risk detection. Therefore, the project focused on advancing STT’s digital transformation strategy and leveraging AI technologies to develop an innovative fraud and corruption risk assessment solution. This technological shift is crucial for ensuring more effective governance and addressing systemic challenges in the management of public and EU funds.

Under the Technical Support Instrument (TSI), the European Commission, in cooperation with the OECD, provided targeted technical assistance to STT between 2024 and 2026. The support focused on strengthening institutional and analytical capacity to address corruption risks in public procurement. Activities included mapping of existing data ecosystems, national stakeholder consultation and engagement, STT data governance and management practices assessment and advice for improvement, iterative development of a proof-of-concept methodology and analytics solution within a secure data environment, and capacity building workshops. The initiative aimed to enhance risk-based supervision mechanisms and reinforce the protection of EU and national public funds.

The project delivered a comprehensive assessment of data governance and management practices, alongside a state-of-the-art model to estimate the probability of fraud risk in public funds procedures. The results also included harmonised and enriched datasets and an ML-supported Proof of Concept (PoC) system with defined technical requirements for further development. The model performs more than twice as well as random guessing - and within the highest 10% of contracts by predicted risk score, it captures roughly five times more known positives than random investigation would yield. Through the model and interactive dashboard, STT can prioritise high-risk cases, support evidence-based investigations, and improve operational efficiency.