AI-Driven Epistemic Threats: Mass Radicalization, Persuasion at Scale, and Plausibility Flooding

This project documents a modern influence-risk shift: generative AI doesn’t just increase the volume of falsehoods — it increases the volume of plausible narratives, delivered in infinite variants, localized to any dialect, and wrapped in synthetic credibility and synthetic consensus. The core risk is not “people fooling themselves with AI.” The step-change is that a small number of operators can manufacture belief and consensus appearance at population scale faster than verification, institutions, and media can respond — with attribution collapse (“no author,” no single origin). We publish: (1) an evidence-backed threat model, (2) a structured “slider” framework (what knobs change impact), (3) indicators defenders can actually monitor (paraphrase swarms, citation laundering chains, coordination signatures), and (4) mitigation mappings across platform, institutional, and model/provider layers. Safety note: This is a defender project. We avoid operational playbooks and focus on detection, measurement, and hardening.