VERITRUST

EU ACADEMIA HACKATHON 2026 • Track 2: Misinformation — Truth and Digital Awareness • PROJECT: "VERITRUST — European Information Trust Ecosystem" • ________________________________________ • 1. EXECUTIVE SUMMARY • VeriTrust is a comprehensive digital ecosystem designed to combat disinformation and misinformation. The project is developed within the framework of 2026 EU policies, including the Digital Services Act (DSA), the Strengthened Disinformation Code of Practice, the European Democracy Action Plan, and the European Media Freedom Act (EMFA). • Core Innovation: A triple synergy of AI + Human + Society. It is not just automation, but a solution aimed at empowering users, increasing media literacy, and ensuring platform transparency. • ________________________________________ • 2. IN-DEPTH PROBLEM ANALYSIS • 2.1. Scale and Statistics • According to Europol data, by the end of 2026, 90% of online content may be AI-generated (Resemble AI Threat Report 2026). • In 2025, 1,567 confirmed deepfake incidents were recorded, resulting in $1.3 billion in confirmed fraud damages. • Ahead of the European Parliament elections, 34 platforms are signatories to the Disinformation Code of Practice. • While EDMO (European Digital Media Observatory) unites over 40 fact-checking organizations, many languages and regions remain underserved. • 2.2. Root Cause Analysis • Technical: Deepfake technologies evolve every 3 months, leaving detection models struggling to keep pace. • Social: "Echo chambers" trap users in content that only confirms their existing biases. • Legal: While DSA and the Disinformation Code exist, monitoring and reporting systems are still maturing. • Educational: Literacy programs are insufficient for youth, and the elderly are often left behind by digital shifts. • Economic: The business model for spreading misinformation (clickbait, engagement-based algorithms) remains largely unchanged. • ________________________________________ • 3. SOLUTION ARCHITECTURE — "VERITRUST" • VeriTrust consists of 5 interconnected components: • 3.1. TRUSTLENS — AI-Powered Real-Time Detection Engine • Theory: Real-time analysis of text using BERT, RoBERTa, and custom transformer models. Visual/video verification via OpenCV and DeepFace. • Credibility Score (0-100): Reliability ratings for every news story/source. • Multi-modal Analysis: Simultaneous analysis of text, images, audio, and metadata. • Cross-Language Detection: Specialized models for all 24 official EU languages. • C2PA Integration: Adherence to Content Provenance and Authenticity standards. • Real-Time API: Browser extensions and bots (Telegram/WhatsApp) for instant user verification. • 3.2. MINDGUARD — Information Literacy and Gamification • Theory: A game-based "pre-bunking" approach inspired by "The Fact-Checking Foxes" (2025 Council of Europe Hackathon winner). • Pre-Bunking Mini-Games: "Fake or Fact?" — Identify news authenticity in 60 seconds. • Echo Chamber Breaker: Based on the "Echobreaker" (Georgia, 2025) concept, exposing users to diverse viewpoints. • Media Literacy Quests: Specialized curricula and quizzes for schools and universities. • Personal Trust Score: Tracking user literacy progress and milestones. • 3.3. TRANSPARENCY HUB — Platform Accountability • Theory: Centralizing the DSA-mandated disclosures from Very Large Online Platforms (VLOPs) regarding moderation and algorithms. • Platform Scorecard: Ranking platforms based on their disinformation combat efforts. • Algorithmic Audit Trail: A system to track content moderation decisions. • Ad Library Integration: Transparency for political ads and their targeting data. • Whistleblower Portal: Anonymous reporting for employees and researchers. • 3.4. FACTNET — EDMO and Fact-Checker Integration • Theory: Aggregating real-time data from EDMO’s 40+ fact-checking organizations. • Live Fact-Check Feed: Real-time stream of articles from EDMO hubs and EFSCN members. • Multilingual Auto-Translation: Using Neural Machine Translation (NMT) for all 24 EU languages. • Collaborative Investigation Workspace: A shared space for researchers and journalists. • 3.5. POLICY WATCH — Monitoring and Analysis • Theory: Tracking legal shifts and policy impacts on the information environment. • DSA Compliance Tracker: Monitoring how platforms meet their legal obligations. • Policy Brief Generator: Automated summaries and recommendations derived from research data. • ________________________________________ • 4. TECHNOLOGY STACK • Frontend: React.js, TypeScript, D3.js (Visualization). • Backend: Python FastAPI, PostgreSQL, Redis, Kafka. • AI/ML: Hugging Face (BERT, DeBERTa), TensorFlow/PyTorch. • Deepfake Detection: OpenCV, DeepFace, FaceForensics++, C2PA SDK. • NLP Pipeline: spaCy, LangChain (Retrieval Augmented Generation - RAG). • Security: OAuth 2.0, GDPR compliance, Zero Retention Mode. • ________________________________________ • 5. EU LEGISLATIVE ALIGNMENT • Document/Initiative • Key Requirement • VeriTrust Response • Digital Services Act (DSA) • Risk assessment and mitigation for VLOPs • Transparency Hub + Algorithmic Audit • Strengthened Disinformation Code • 44 commitments / 127 specific measures • Platform Scorecard + Crisis Tracker • European Democracy Action Plan • Safeguarding elections • Pre-Bunking Games + Live Fact-Check Feed • AI Act • Transparency and Human Rights • Explainable AI + C2PA Tracking • ________________________________________ • 6. EXPECTED IMPACT (KPIs) • Engagement: 50,000+ active users within the first 6 months. • Accuracy: 85%+ Precision and 90%+ Recall in fake news detection. • Education: 10,000+ students trained via MindGuard games. • Coverage: Localization and integration across 15+ EU member states. • ________________________________________ • 7. HACKATHON EXECUTION PLAN (12 HOURS) • 0-2h: Problem analysis, EU legislation review, and architecture finalization. • 2-4h: TrustLens MVP: BERT-based text analysis prototype. • 4-6h: MindGuard MVP: Development of 2 mini-games + UI. • 6-8h: Transparency Hub: Platform rating dashboard and visualization. • 8-10h: FactNet: Integration of live EDMO data feeds. • 10-12h: Demo video, presentation slides, and pitch rehearsal. • ________________________________________ • 8. COMPETITIVE ADVANTAGE • Unlike existing tools (NewsGuard, InVID, etc.), VeriTrust is a holistic ecosystem. It combines detection with education, legal compliance, and platform transparency. It is specifically tailored for the 24-language European landscape and operates as an Open Source project to foster community trust. • ________________________________________ • 9. CONCLUSION • VeriTrust is not just an AI detector; it is a bridge of trust between technology, policy, and society. Our next steps involve presenting the MVP at the hackathon and seeking partnerships with EDMO and Horizon Europe to scale this into a pan-European reality.

Built With

• deepface
• kimi
• langchainki
• opencv
• pytorch
• tensorflow