The publication, Review of Digital Transformation Opportunities and Challenges for Corruption, examines how artificial intelligence (AI) and other digital solutions can be used to identify corruption risks, enhance transparency, and automate certain control procedures. At the same time, the authors consider the limitations of these technologies and the new vulnerabilities that may arise when they are implemented.

The report is based on a review of 38 studies**. The authors analyse digital transformation through six interconnected domains: people, data, software, hardware infrastructure, processes, and communication. This approach makes it possible to assess not only the algorithms themselves, but also the conditions that determine their practical effectiveness.

One of the central findings is that AI is not a ready-made universal solution for fighting corruption. Its application may be useful, for example, for anomaly detection, analysis of procurement, financial transactions, declarations, or other datasets. However, the results depend on the quality of the underlying data, the level of digital infrastructure, the institutional environment, users’ skills, and the existence of clear oversight procedures.

The most developed area in the existing literature remains software solutions: algorithms, platforms, and machine-learning models. At the same time, much less attention is paid to infrastructure, organisational processes, and the human factor. The authors note that poor connectivity, unstable energy supply, insufficient computing capacity, and low digital literacy can significantly limit the use of AI, especially in countries with less developed digital environments.

A separate problem concerns data. AI requires complete, high-quality, and comparable datasets, whereas corruption-related data are often fragmented, incomplete, or based on indirect indicators. In addition, corrupt practices are usually hidden, which makes it difficult for researchers to build a reliable dataset of confirmed cases on which models can be trained.

The authors also draw attention to the risks of opacity and algorithmic bias. If a model operates as a “black box,” users may find it difficult to understand why a particular case has been classified as high-risk. If a system is trained on incomplete or distorted data, it may reproduce existing errors and institutional biases.

The report also stresses that evidence of AI’s direct impact on reducing corruption remains limited. Positive results from the use of digital tools have been recorded in a number of countries, but it is not always clear whether these results reflect an actual reduction in corruption or more effective detection of potential violations.

In this context, the authors propose that AI tools should be viewed primarily as decision-support systems rather than as a replacement for human oversight. They can help identify suspicious links and transactions more quickly, but the final assessment should remain with specialists.

*BridgeGap is a four-year international research project running from January 2024 to December 2027 under the Horizon Europe programme. It builds on the earlier EU-funded anti-corruption project ANTICORRP and aims to advance an interdisciplinary understanding of corruption, expand knowledge and data on political corruption and other forms of undue influence, and analyse the potential use of advanced technologies to detect, prevent, and fight corruption.

**To select the sources, the authors used academic search engines and databases, including Google Scholar, Web of Knowledge, JSTOR, MEDLINE, and ScienceDirect, as well as Elicit, an AI tool for searching and systematising academic literature based on Semantic Scholar and OpenAlex. The results obtained with Elicit were additionally checked manually, since such tools may make errors when interpreting sources and extracting data.