Let me give you a number that's hard to process: corruption in the EU costs up to EUR 990 billion per year, according to the European Parliament. Public procurement, with its EUR 2 trillion annual spend, is one of the most corruption-prone areas.
The European Public Prosecutor's Office (EPPO) closed 2025 with 3,602 active investigations and estimated damage to public finances exceeding EUR 67 billion. They opened 2,030 new cases in 2025 — 35% more than the previous year. Judges issued freezing orders totaling EUR 1.13 billion.
These aren't abstract numbers. This is public money — taxpayer money — being diverted, wasted, or stolen through manipulated procurement processes.
How procurement fraud actually works
The Hollywood version of fraud is dramatic: briefcases of cash, shadowy meetings. Real procurement fraud is more mundane and harder to spot.
Bid rigging remains the most common form. In 2023, it accounted for 44% of all cartel infringement decisions across the EU. It comes in a few flavors:
Cover bidding: Companies submit deliberately uncompetitive bids to create the illusion of competition while ensuring a pre-selected winner. The prices look different, but the outcome was agreed in advance.
Bid rotation: Companies take turns winning tenders in a geographic or temporal pattern. Company A wins this quarter, Company B wins next quarter. Over time, everyone gets their share. The procuring authority might never notice because each individual tender looks competitive.
Bid suppression: Companies agree not to bid, or to withdraw their bids, reducing competition for a designated winner. When a tender consistently attracts one or two bidders, this might be why.
Market allocation: Competitors divide the market — by region, by sector, by client — and agree not to compete in each other's territory.
Specification manipulation is the other big one. A procurement officer writes requirements tailored to a specific supplier. Not obviously — but through technical details, qualification thresholds, or reference requirements that only one company can meet. From the outside, the tender looks open. In practice, the winner was chosen before publication.
Why it's hard to catch
The fundamental problem is information asymmetry. A procurement officer reviewing five bids sees five independent proposals. They might not know that three bidders share a parent company, or that the pricing patterns suggest coordination, or that the winning bidder has won suspiciously often in a specific region.
Even well-intentioned procurement teams can miss the signs because detection requires analyzing patterns across many tenders, not just evaluating a single one. You need to look at who bids on what, how often, at what prices, and whether the patterns make statistical sense.
What's changing: AI-powered detection
Several countries are now using AI specifically for fraud detection in procurement, and the results are worth paying attention to.
Spain's competition authority (CNMC) launched BRAVA in 2024 — an AI tool that classifies bids as potentially collusive or competitive using machine learning algorithms trained on the national procurement database. It analyzes bidding patterns at a scale no human team could match.
The UK's Competition and Markets Authority (CMA) announced a trial of an AI collusion detection tool in January 2025. The pilot with one government department already showed results in flagging suspicious bidding patterns.
Brazil's ALICE system has been the poster child for AI in procurement oversight. It suspended R$9.7 billion (roughly EUR 1.7 billion) in suspicious bids and cut audit time from 400 days to 8 days.
Ukraine's Dozorro system achieved a 298% increase in collusion detection rates across 3,000+ daily tenders — while the country was in the middle of a war, which says something about the prioritization of procurement integrity.
The OECD updated its Guidelines for Fighting Bid Rigging in Public Procurement in June 2025, and launched two EU-funded projects: one helping Austria, Bulgaria, Croatia, Cyprus, Greece, and Romania (started September 2024), and another helping Czechia, France, Ireland, Latvia, Poland, and Portugal (started 2025). Latvia being in that second group is notable — there's an acknowledgment that more needs to be done.
Transparency as prevention
Detection matters. But the best fraud deterrent isn't catching it after the fact — it's making it harder to commit in the first place.
This is where thorough, evidence-based bid evaluation changes the equation. When every finding in a procurement evaluation is documented with specific quotes from the documents — when the entire decision trail is traceable — it becomes much harder to justify a rigged outcome.
If a tailored specification excludes legitimate competitors, an AI that reads and evaluates all bids against all criteria will produce findings showing that the winning bid wasn't actually the strongest on merit. The paper trail exists. The evidence is there.
If bid prices are coordinated, an AI that reads pricing proposals in detail and compares them against market data and technical scope can flag inconsistencies that manual review might miss.
The point isn't that AI replaces human judgment in these situations. It's that AI creates a complete, auditable record of what the documents say. You can't bury a finding in page 180 when something has read and documented page 180.
What this means for procurement teams
If you're running a clean procurement operation — and most teams are — thorough AI evaluation is your shield. It protects your decisions with evidence. It makes your process defensible against legal challenges. It demonstrates due diligence in a way that manual review can only partially achieve.
And if there are problems in the system — whether from external fraud or internal pressure — the evidence trail makes them visible.
EUR 29.6 billion. That's the estimated cost of corruption risk in EU procurement between 2016 and 2021, according to EU-funded research. Even reducing that by a fraction would fund a lot of schools, hospitals, and infrastructure.
The tools to fight this exist now. Using them is a choice.