Generative AI has armed hostile actors with unprecedented power to create convincing fakes and disinformation. These narratives, cloaked behind botnets, encrypted chats, and decentralised platforms, often spill into real-world crime.

In 2024, a coordinated disinformation campaign slipped past Romania’s defences, forcing the annulment of a national election; street violence erupted soon after.

European law enforcement agencies urgently need a tool to anticipate and counter these threats.

The Fake News Risk Mitigator (FERMI) – launched as an EU-project in July 2022 – gives law-enforcement agencies a real-time early-warning dashboard built around three stages:

 

• Step 1: Classify the content

 

Bot-or-not: Is the content AI-generated or automated?

Sentiment: Is the tone angry, fearful, or reassuring?

Intent: Is the author trying to inform, manipulate, recruit, or incite?

 

• Step 2: Map the network

 

Who: Which accounts are linked to a suspect post?

Influence: Who are the key amplifiers, sources, victims and likely targets?

Vulnerability: Which communities are most vulnerable to the narrative?

 

• Step 3: Predict & alert

 

Real-world threat: What’s the probability that a rumour could escalate into criminal activity, such as hate crimes, riots, or other offences?

Alerts: Is an instant warning to police necessary? If so, FERMI alerts the police so they can quickly counter-message, reach out or deploy on the ground before trouble starts.

 

We led the development of the spread-analysis engine, user interface, quality control, and risk management.

FERMI is now being rolled out EU-wide. Three successful projects were completed during the pilot phase, using live data with operational police teams.

 

• Finland

 

The Ministry of the Interior mapped bot-driven Russian refugee narratives within hours, pinpointed high-influence accounts, and helped officers focus outreach and patrols before extremist violence could ignite.

 

• Belgium (Eurotop summit)

 

The Federal Police used real-time alerts to redeploy units to likely clash points, run a rapid fact-check campaign, and stop far-left and far-right groups from clashing violently around the summit venue.

 

• Germany (Munich Security Conference)

 

Bavarian police tested FERMI’s full toolset against a left-wing extremist scenario, using crime-flow forecasts to harden vulnerable sites and coordinate messaging.