Over the past decade, the European Union has funded a long list of tools meant to detect and counter disinformation: software that flags manipulated media, systems that trace coordinated influence campaigns, platforms that help fact checkers work faster. What it has lacked is a clear answer to a basic question. Which of these tools actually work when they leave the lab and land on the desk of a journalist or security analyst?

A new Horizon Europe project coordinated by Johannes Gutenberg University Mainz (JGU) in Germany aims to find out. Called FIMI-RESIST, it is led by information systems researcher Prof. Vera Schmitt, head of JGU’s XplaiNLP Research Group, and brings together ten partners from nine countries. The project will run for 30 months with approximately โ‚ฌ3.5 million in funding.

Not Another Tool

FIMI stands for foreign information manipulation and interference, a term European institutions use for coordinated efforts by state and non-state actors abroad to distort public debate, often through deceptive or covert means. Such campaigns have been documented around elections, the war in Ukraine and public health crises, and they increasingly use automated accounts and AI-generated content.

Unlike many earlier projects, FIMI-RESIST will not build another stand-alone tool. Instead, it will develop a shared framework for evaluating and harmonizing the solutions Europe already has. The consortium plans to assess 20 tools previously funded by the EU using common criteria and shared benchmarks, so that their strengths and weaknesses can be compared directly rather than judged by each developer’s own measures.

The aim is practical: to identify which approaches suit which users, operational settings and threat scenarios. A tool that performs well on a benchmark dataset may be of little use to a newsroom working against a deadline, or to an analyst who needs to understand why a system flagged a piece of content before acting on it.

Testing in the Real World

To get at that question, the project will work directly with journalists, fact checkers, security stakeholders and civil society organizations. The partners will test how well the tools fit into these groups’ daily workflows and where gaps remain.

The consortium reflects that mix. Alongside JGU’s XplaiNLP Group, it includes the Kempelen Institute of Intelligent Technologies (KInIT) in Slovakia, the Centre for Research and Technology Hellas (CERTH) in Greece, the University of Amsterdam, Logically/INSIKT in the United Kingdom and Spain, Factiverse in Norway, Exorde Labs in France, the Cyfluence Research Center and the broadcaster Deutsche Welle in Germany, and the fact-checking organization VoxUkraine. Gretchen AI, based in Germany, is an associated partner.

Schmitt’s group focuses on explainable natural language processing, the branch of AI concerned with making language-analysis systems transparent about how they reach their conclusions. That emphasis matters in disinformation work, where users must be able to trust and justify a tool’s judgment before acting on it.

Lasting Beyond the Grant

Research projects often produce promising prototypes that disappear when funding ends. FIMI-RESIST plans to counter that by setting up a help desk, a European hub and training programs so that validated methods remain available after the project concludes.

The findings are also meant to inform European policy. The EU has adopted several major laws touching on online information in recent years, including the Digital Services Act, which places obligations on large platforms to address systemic risks; the AI Act, which regulates artificial intelligence systems; and the European Media Freedom Act. Evidence about which countermeasures work, and under what conditions, could shape how those rules are applied.

The project is at its start, so there are no results yet. Its premise, though, reflects a broader shift in the field: after years of building defenses against disinformation, the next step is measuring whether they hold.


Endnotes

  1. Johannes Gutenberg University Mainz (October 7, 2026). “Disinformation and influence operations: Which countermeasures actually work?” EurekAlert! https://www.eurekalert.org/news-releases/1146921

Leave a Reply

Trending