FBK to Develop Core AI Tools for ECLIPSE Project Against Hate Speech

As the technical coordinator of the ECLIPSE project, Fondazione Bruno Kessler (FBK) will develop and provide a suite of advanced and trustworthy Artificial Intelligence tools. These tools will form the technological foundation for the project’s mission to combat online and offline hate speech and disinformation in Europe.

The tools are designed to perform specific, interconnected functions within the broader ECLIPSE Toolbox. In particular, these technologies will be developed based on insights from theory, policy and practice — ranging from social, behavioural, criminological, and legal sciences — to applied expertise from universities, foundations and think tanks, law enforcement authorities, NGOs, and international organisations. This article provides an overview of the ECLIPSE Toolbox components and explains how they work together to address hate speech and mis/disinformation.

Hate speech and mis/disinformation represent growing threats to European values, individual well-being, and social cohesion, eroding fundamental principles such as dignity, freedom, and equality. These phenomena are deeply interconnected: hate speech often thrives on and amplifies mis- and disinformation, creating a self-reinforcing cycle that can escalate into discrimination, polarisation, and even violence [1]. Both share common harmful dynamics, including the normalisation of violence, the manipulation of fear and emotions, and the exploitation of the viral logic of social media platforms. Yet, despite their close relationship, EU policies and initiatives frequently address hate speech and mis/disinformation in isolation, limiting the effectiveness of current responses.

The EU Agency for Fundamental Rights has emphasised the need to reassess the combined impact of these phenomena, particularly on vulnerable communities, and to develop practical, evidence-based approaches that address their manifestation in both digital and physical spaces, an urgency further underscored by recent crises such as the COVID-19 pandemic and ongoing global conflicts [2]. Effectively countering these evolving threats, therefore, requires integrated strategies that operate across online and offline environments [3], while promoting solidarity [4], strengthening content moderation practices, and empowering cyber-activism [5]. In this context, there is a clear need for innovative solutions, particularly advanced tools that bring together artificial intelligence and the social sciences, and which can help bridge the gap between research and practice. Such tools are essential for understanding the complex links between online hate speech and mis/disinformation and for enhancing training initiatives and awareness-raising efforts among European citizens and practitioners.

To meet these challenges, the FBK Team brings together expertise in computer science, physics, and human-computer interaction to develop a lawful and integrated set of tools for detecting, reporting, and analysing hate speech and mis/disinformation. This technical expertise is complemented by a broader multidisciplinary team with knowledge in social sciences, ethics, and practice, alongside long-lasting collaborations with NGOs and civil society organizations, ensuring that the development process actively addresses ethical considerations, potential risks, and real-world implications. These tools form the ECLIPSE Toolbox, a unified framework designed to translate interdisciplinary research into practical impact. The Toolbox comprises four core components: perception tools for identifying harmful content, social network tools for analysing its spread and dynamics, counteraction tools to support mitigation and reporting, and an Augmented/Extended Reality (AR/XR) gamification tool to enhance training and awareness. Keep reading to learn more about the ECLIPSE Toolbox components. While each tool description highlights the partners directly responsible for its technical development, it’s important to note that all components are grounded in research conducted collaboratively with university and research partners across the consortium, and have been verified and validated by our legal and ethical partners to ensure compliance and responsible deployment.

The Perceptions Tool

An advanced Natural Language Processing (NLP) system for detecting and analysing hate speech and disinformation across multiple languages (English, French, German, Italian). It will perform emotion detection and topic modelling on large datasets from social media and news platforms. This tool will be developed through a collaboration between The Institute for Artificial Intelligence Research and Development of Serbia (IVI) and Language and Dialogue Technologies (LanD) FBK unit.

The Social Networks Tool

A Social Network Analysis (SNA) component designed to map relationships between online accounts. Its purpose is to identify key influencers, detect coordinated behavior, and analyse how harmful content spreads through digital communities. This tool will be developed through a collaboration between the Sabancı University (SU) and the FBK Complex Human Behaviour Lab (CHuB).

The Counteraction Tool

A Natural Language Generation (NLG) system that will generate context-sensitive counter-narratives to hateful or false content. This tool will use a Human-in-the-Loop (HITL) approach, where law enforcement and civil society end-users will help guide and refine the AI-generated responses. The LanD unit at FBK will lead the development of this tool, working in close collaboration with Law Enforcement Authorities (LEAs), Civil Society Organisations (CSOs), and international partners within the ECLIPSE consortium.

The AR/XR Gamification Tool

This training and awareness tool will integrate gamification and digital storytelling with Augmented and Extended Reality (AR/XR) technologies. It will be developed by the FBK MODIS unit in collaboration with the partner CREW, which brings expertise in creating immersive, human-centred performance experiences. This tool will provide interactive training simulations for professionals.