MMOOZZGG.com

AI against corruption

Arachne risk scoring and EPPO: how the EU protects its budget

6 min read2 sources

Arachne risk scoring helps the EU find risky projects among recipients of shared money: the system brings together data on projects, their owners and contractors and calculates risk indicators. Since 2026, the new Arachne+ version has run with 78 indicators instead of 107. The European Public Prosecutor’s Office (EPPO) handles criminal cases of fraud with the EU budget, and its statistics show where the damage is largest.

On this page · 6 sections

In short

Arachne is a European Commission system that looks for signs of fraud, conflicts of interest and irregularities with EU budget money. The new Arachne+ version calculates 78 risk indicators from data on projects, company owners and officials, but makes no decisions itself. The European Public Prosecutor’s Office (EPPO) investigates the cases: at the end of 2024 it had 2,666 active investigations with damage of about 24.8 billion euros.

Arachne: what it checks

The European Commission describes Arachne+ as a corporate data-mining and risk-scoring tool. Its task is to help detect and prevent fraud, conflicts of interest and irregularities with EU budget money.

The new version covers all EU funds and all management modes: shared with member states, direct and indirect. The system flags anomalies in four areas: procurement, eligibility for funding, reputational risk and the concentration of funds with the same recipients.

Arachne+ has 78 risk indicators; the old version had 107. The European Commission calls the new set streamlined and targeted: there are fewer indicators, and each is tied to a specific risk.

The transition was short. Arachne+ went live on 27 April 2026, users moved to it until 30 June, and the old system is retired by the end of June 2026.

Data on owners and officials

Arachne’s strength is that it looks not only at the project but at the people behind it. Data on projects and recipients comes from EU institutions and member states, and the system takes external data from the Orbis and World Compliance databases through the vendor Moody’s Analytics.

External databases provide company owners, sanctions lists and politically exposed persons. For example, this shows that a project contractor belongs to the same person as the grant recipient, or that an official who influences the choice is among the owners.

This is a check that is hard to do by hand: a project can have many contractors, each with its own chain of owners in different countries. How to find such links is covered on the page on conflicts of interest.

From the 2028–2034 budget period, EU member states will be required to send data to Arachne+. For now its use is voluntary, and the European Commission will assess readiness for the mandatory regime by the end of 2027.

Why the system does not decide by itself

The European Commission states it plainly: Arachne+ supports risk-based decisions but makes no automatic decisions. The tool is designed only as support for whoever decides.

A high risk score means the project deserves a closer look. It does not mean the recipient is guilty: a link to an official may be disclosed and lawful, and a concentration of funds may have a market explanation.

The decision stays with the body that manages the program, and if there are signs of a crime, with the prosecutor and the court. The machine narrows the search, and a person answers for the conclusion.

The legal basis is written down too. Article 36(6) of EU Financial Regulation 2024/2509 provides for a corporate data-mining and risk-scoring tool, and the Joint Declaration of 7 December 2023 requires a review of its use by the end of 2027.

EPPO: the scale of investigations

The European Public Prosecutor’s Office (EPPO) started work in June 2021 and investigates crimes against the EU’s financial interests. At the end of 2024 it had 2,666 active investigations, with estimated damage to the EU budget of about 24.8 billion euros.

Over the year, the number of active investigations grew by 38% and the damage estimate by 22.5%. In 2024, EPPO opened 1,504 new investigations, 9.7% more than a year earlier, with damage of 13.07 billion euros.

Cases come from different places. In 2024, EPPO processed 6,547 crime reports, 56% more than a year earlier: more than 70% from private persons and organizations, about 27% from national authorities and only 1.7% from EU institutions.

EPPO sees the last figure as a weak point and calls on EU institutions to detect and report suspicions more actively. For any risk system, the lesson is the same: a signal that nobody passes to an investigator does not protect the budget.

The result shows in court and in money: 205 indictments, 47% more than in 2023, and asset freezes of 849 million euros. Separately, EPPO runs 311 cases on the NextGenerationEU program, with damage of 2.8 billion euros.

The VAT share of the damage

VAT cases are 18% of EPPO’s active investigations, 488 cases, but 53% of all damage, about 13.15 billion euros.

EPPO links this share to cross-border VAT fraud, which almost always involves criminal organizations. The signs of such schemes show only when data from several countries is matched: supplies between companies in different states, deductions and payments must be viewed together.

The lesson for audit work: a small share of cases can produce most of the damage. So a risk system should rank signals by the amount at stake, not only by the number of violations.

How the signs of tax schemes show in data and which matches this needs is covered on the page on VAT schemes.

What Russia can take from this

The EU experience is useful to Russian public bodies and state-owned companies that distribute subsidies, grants and contracts. Five decisions worth adopting:

1. Match data on owners with data on money recipients. Arachne shows that the main risk often lies not in the project document but in who owns the contractor.

2. Publish the set of indicators and review it based on experience. The European Commission cut the set from 107 to 78 and called it targeted.

3. Write into the rules that the system does not decide by itself. This protects both those audited and the auditors from an algorithm’s error.

4. Rank signals by damage. EPPO’s VAT experience shows that 18% of cases can produce 53% of the damage.

5. Introduce mandatory data sharing step by step. In the EU, the use of Arachne+ is voluntary for now, and mandatory data sharing will start in 2028 after a readiness assessment.

Other countries and their systems are collected on the global experience page.

Sources and fact-checking

Checked by the editors on 16 October 2026.

  1. European Commission, Anti-Fraud Knowledge Centre (checked 16 October 2026), 16 October 2026 (antifraud-knowledge-centre.ec.europa.eu)
  2. EPPO, Annual Report 2024, 3 March 2025 (eppo.europa.eu)

Found a mistake? Tell us

Publications

Mike Mozg

Agentic AI use cases at AT&T: nearly 500 processes now run on agents

AT&T no longer uses AI only to help staff work faster: it is rebuilding whole business processes around agents. SiliconANGLE reported on 14 November 2026 that nearly 500 agentic AI use cases at AT&T are already in production. Here is how the company rebuilds them, where a human still checks the agent, and what other large companies can take from it.

News4 min

For government bodies

We build such systems for government bodies and state companies, from data analysis to a pilot on one scheme. The data stays on the organization's own servers.

Contact us