Recommending that money comes back is not the same as getting it back, and EU fraud enforcement lives in the gap between the two. The European Anti-Fraud Office recommended the recovery of 597 million euro in its 2025 annual report, closed 209 investigations and opened 254 new ones. Every euro of that figure now depends on someone else acting.
OLAF investigates administratively. It cannot prosecute, cannot seize assets and cannot compel a national authority to follow its recommendations. It writes a file, sends it to a managing authority or a prosecutor, and waits. The office reports recovery rates above 96 percent where follow-up has been completed, a strong number that quietly concedes how much follow-up remains incomplete.
The decade-long total tells the same story from the other end. Roughly 6.8 billion euro has been retrieved or protected since 2015, which averages under 700 million a year against an EU budget running past 190 billion annually plus the recovery facility on top. Detected fraud is a fraction of estimated fraud, and recovered fraud is a fraction of detected fraud.
Two structural fixes are moving. The first is the European Public Prosecutor’s Office, which took over criminal investigation of budget offences in participating states and has grown faster than anyone forecast. OLAF and the EPPO signed a working arrangement in 2021 that divides the ground. The prosecutor handles crime, the office handles irregularities, and each refers to the other. In practice the boundary blurs, because most fraud files begin as irregularities and only later look criminal.
The second fix targets the largest single leak. Member states agreed in May 2026 to give the EPPO and OLAF direct access to cross-border VAT data held by Eurofisc. Carousel fraud costs national treasuries far more than misspent grants ever will, and investigators have complained for years that they could see the pattern only after the traders had dissolved. Direct data access shortens that lag.
Neither fix touches the deeper problem, which is that recovery happens in national courts under national procedure. A Commission recommendation lands in a system with its own backlog, its own evidentiary standards and its own view of how urgent EU money is. Cases involving structural funds routinely take longer than the programming period that disbursed them.
The Commission signalled in July 2026 that it wants to tighten the framework further, and a communication this year could produce legislative proposals on cooperation between OLAF, the EPPO and national authorities. The obvious candidate is a rule obliging member states to report back on what they did with a recommendation, within a deadline.
Governments will resist that, and the resistance is not purely defensive. Prosecutorial independence is a real principle, and an administrative office setting deadlines for national prosecutors would test it. The counterargument is that the money belongs to the Union, not to the state that happened to spend it.
This year’s report added something new: recommended actions aimed at stakeholders rather than findings aimed at files. It is a small shift from investigation toward prevention, and it reflects an office that has concluded it cannot chase its way out of the problem. Preventing 18 million euro from being spent, as OLAF did last year, costs far less than recovering it afterwards.
The next EU budget will multiply the exposure, with performance-based disbursement moving large sums against milestones that are easier to claim than to verify. An anti-fraud system that already struggles to close its existing files will be asked to police a larger and looser one. The 96 percent recovery rate is worth defending. Whether it survives the scale-up is a different question.
