Detecting disinformation in AVALANCHE

From trusted knowledge to verified truth

Online disinformation poses an increasing threat to public trust, democratic stability, and social cohesion. Law Enforcement Agencies (LEAs) and analysts are often faced with a vast and fast-changing digital landscape, where misleading narratives spread rapidly across news outlets, blogs, and social media. Manual verification is time-consuming and often fragmented, making it difficult to respond effectively.

To address this challenge, AVALANCHE employs a two-stage disinformation detection framework that combines explainable artificial intelligence (XAI), data integrity verification, and multilingual reasoning. The system is designed to support Law Enforcement Agencies (LEAs) in assessing the credibility of online information, from mainstream news and social media to OSINT sources, in a transparent and evidence-based way.

Stage 1: Building a Trusted Corpus

The process begins with the creation of a continuously updated Trusted Corpus — a repository of reliable information drawn from authoritative sources such as verified media outlets, government portals, and institutional data providers. This corpus functions as a living knowledge base, allowing analysts to trace the origin and evolution of factual narratives.

It is designed to reflect reality as accurately as possible, ensuring that future assessments of online content are grounded in credible, verifiable information rather than isolated claims.

Stage 2: Detecting and classifying disinformation

When a new article, post, or claim is analysed, AVALANCHE compares it against the Trusted Corpus to uncover relevant context and potential evidence. The system uses advanced reasoning and explainable AI (XAI) methods to evaluate whether the information aligns with or contradicts established facts. The outcome is presented as a credibility assessment, for example, Likely True, Likely False, or Uncertain, accompanied by clear, human-readable explanations that allow investigators and analysts to understand the reasoning behind each result.

By combining human expertise with AI-driven verification, AVALANCHE offers a balanced approach to combating online disinformation. Every decision is traceable, every source is transparent, and every conclusion can be revisited.