CENTRA

SafeCall is an AI-assisted voice-phishing intervention prototype designed to help people recognize manipulation and stop risky financial actions before money is lost. Scammers often impersonate banks, prosecutors, or delivery services. They gradually create urgency, fear, secrecy, and pressure to transfer money. A simple keyword-based warning is not enough to address these evolving conversations. SafeCall demonstrates an end-to-end protective experience: - Synthetic English voice calls with two 20-turn phishing scenarios. - Escalating risk assessments with evidence-based explanations. - Meta AI integration for speech transcription and contextual analysis. - A smart-glasses-style browser HUD showing timely, concise warnings. - Simulated GPS movement toward a bank. - A fictional SafeBank application that holds risky transfers. - Post-incident safety guidance and independent verification steps. The team implemented Meta Model API integration and performed API calls during development. The reproducible demonstration also supports local synthetic-data replay to keep the experience reliable without requiring live API calls for every presentation. The smart-glasses interface, bank proximity, and banking application are simulations, not integrations with physical Meta glasses, actual financial institutions, or real phone-call interception. Our vision is to move scam protection beyond passive detection toward timely, understandable intervention at the moment a person is being pressured to act.