Opportunities in detecting fraud and corruption with EU funds

A recent study by researchers from the Centre for Strategy & Evaluation Services (CSES) and the Utrecht University School of Economics (U.S.E.) has explored technological developments in budgetary control within the EU. The misuse of EU funds remains a serious problem and forms a threat to public confidence in the EU. Data technology can potentially facilitate and speed up the detection of fraud and corruption. ‘There is plenty of data and new technologies available in all member states,’ says Vitezslav Titl of U.S.E. ‘Researchers use them, NGO’s use them and within the member states, the audit offices sometime use them. We know they work, but they are underused at the EU level. There are missed opportunities in detecting fraud and corruption, which could be done in a much more efficient way and saving lots of resources.’

The study contains an overview of new developments that can be expected and how these could be used to protect the EU’s budget by preventing fraud and corruption with EU funds in the future. ‘At the moment investigation officers often are wasting their time finding problems that can automatically be detected,’ Vitezslav Titl continues. ‘Automation can overcome very simple tasks, which leaves a lot of potential to do the actual investigation. The data are there, they just have to be connected. The quality often can be improved but already now we have lots of data that is not used well. Using available data and technologies at the EU level could not only help finding corruption and fraud, but also doing this more efficiently. And the officers can focus on the investigative work itself not simple tasks they sometimes now spend time on. This is especially interesting from the point of view of the net payers (and budgetary critical members) like the Netherlands or Germany.’

Using available data and technologies at the EU level could not only help finding corruption and fraud, but also doing this more efficiently.

Protecting the EU budget

The misuse of EU funds remains a serious problem, with Member States reporting a total of 12,455 irregularities, amounting to EUR 1.77 billion, in 2022. The misuse poses a serious threat to the EU's ability to advance its strategic priorities and maintain public confidence in the EU. To address this, digitalisation is at the heart of the strategic vision of the European Commission and other bodies responsible for management and control of EU expenditure. Budgetary authorities have increasingly used new digital technologies to protect the EU budget.

Serious threat to public confidence in the EU

Examples of digital tools operated by the European Commission to help identify risks of irregularities are Arachne (a risk-scoring tool used by managing authorities (MAs) to detect risks of fraud and irregularities in the use of European Structural and Investment Funds), the Early Detection and Exclusion System (EDES) (a database allowing EU bodies to flag financial risks posed by (potential) recipients of EU funds) and the Irregularity Management System (IMS) (a database within which Member States report irregularities in the management of EU funds.)

‘Those are different tools, based on ‘economic intuition’, Titl states. ‘Quite simple calculations that show that there potentially is a problem. For example, on the calculation of unique winners in procurement contracts. In Hungary for instance, projects in some cities are done by always the same company in 60-70% of the cases (a company with connections to the ruling Fidesz party). That stands out. But also, in general, you could look at the share of procured work in contrast with directly awarded work (no competition) in these data.’ 

Possible future developments

But there is scope for further application of digital technologies. New technologies that could be applied, include AI-powered systems, machine learning, large language models, big data, robotic process automation and satellite imagery. Many of these technologies are inter-connected.

Harnessing the power of AI tot detect fraud

These new technologies could increase the efficiency and quality of budgetary control. AI and machine learning algorithms are proving accurate in detecting potential risk or cases of fraudulent spending and corruption. Machine learning can also be used to automate checks on operations in public procurement and for real-time monitoring of spending.

‘Machine learning recognizes patterns in the data faster than we as humans can. If we have a sample of corrupt or fraudulent subsidy cases, an AI tool can learn to detect it within the sample that we have and also predict for new samples. It could check whether it looks like there could be a problem,’ says Titl, ‘and consequently, that case could be investigated further.’

Challenges in the use of new technologies include the need for uniform data collection, interoperability of data and systems, cost, privacy regulation compliance, ethical concerns relating to biases embedded in AI systems and a high number of false positives.

Inspiration at different levels

The study by CSES and U.S.E. has assessed the advantages and limitations of these technologies, as well as factors (including data privacy, legal requirements, technical and cost issues) that facilitate or hinder their uptake and successful use for EU funds under different management modes.

Integration at EU level still lacking

‘Different countries try different things on their own,’ Titl explains. ‘Different technologies, machine learning etc. show what can be done. It could be inspiration for others and for EU level. But this ‘second step’ is missing so far; it seems to be difficult to integrate this at EU level.

The inspiration comes from different levels. In some countries, it is the government that is developing a tool to detect corruption and fraud (in Italy for instance), but it can also be civic movements and NGO’s (for instance in Hungary and Czechia). In Hungary, they try to prevent corruption, also because there are lots of worries about that, and the government is probably part of the problem.’

Obstacles in cooperation

To date, a broad and consistent deployment of data-driven technologies in budgetary control across the EU has remained limited due to differences in national control strategies and systems, regulatory frameworks, investment capacity, digital competences and political priorities between Member States.

‘At the European level, there are many databases that could be used. For example Irregularity Management System, the tool that at EU level is developed to collect data about frauds contains perfect training data. We also have data on all subsidies. These two datasets could be merged and AI could be used to detect fraud. Certain countries opt out from using these and similar systems such as Arachne. Per se that is okay, but it creates obstacles in cooperation. The datasets are then not compatible. The unification of standards for data is a big issue,’ says Titl.

By opting out, countries hinder cooperation

‘There are also huge differences between countries in terms of understanding what privacy is,’ he goes on. ‘Whenever we suggested to use AI during a discussion with members of the parliament, the German MP’s for example are very hesitant. In Czechia a lot of data that in Germany cannot be published are simply online, in Estonia or Finland the same.

Another example is on the regulation for donations to political parties. In the Netherlands, you don’t have to report donations below 10.000 euro (as in many countries), so you don’t see these on the lists parties provide, because that people or firms that do not want to be seen donate a slight bit less. The rule makes sense, nations can choose their own, and I completely understand that. But on the other hand, in Czechia we have all the information and could use that to predict problems, and identify possible fraud and corruptions, whereas in The Netherlands this is not possible because these donations are not in the data. In Belgium, this is even forbidden. In short: the rules are so different that it is difficult to do something in a harmonized way.

The EU and the member states are trapped in ‘we need to solve this stuff now’ instead of investing and then getting more efficient. This is one of the roles of our report: to show that there are big opportunities missed in terms of how efficient you could be in finding cases of fraud and corruption with European money.

Recommendations

The research team has formulated eight recommendations for the EU committee on budgetary control:

• Continue to enhance existing EU tools for budgetary control
• Promote awareness of and training in the use of existing EU tools for budgetary control
• Consider making the use of EU tools mandatory
• The EU and its Member States could consider pilot projects to explore the possibilities for applying new data-driven technologies to budgetary control
• Support mutual learning, the sharing of good practices and exchanges of information between relevant authorities
• The EU could consider defining common standards for the use of new technologies in budgetary control accompanied by a code of conduct for the proper and ‘fair’ deployment of these technologies for budgetary control
• Assess the costs and benefits before deploying new technologies
• Carry out regular “horizon scanning” to identify potential new technological developments suited for application to budgetary control and share information about such developments with budgetary authorities at EU level and in the Member States.