Learn/Core Concept How does knowledge poisoning work? Knowledge poisoning inserts misleading data into a retrieval store, steering answers when matching chunks reach the model. Validate sources, track provenance and filter writes, as the RAG security guide shows. Data poisoningProvenance |