The GenAI Data Security Initiative for AI & Agentic applications is a focused workstream within the OWASP GenAI Security Project dedicated to securing data across the full lifecycle of generative and agentic AI systems. It addresses the unique security challenges introduced when LLMs, RAG pipelines, embeddings, vector stores, tool integrations, and autonomous agents aggregate and act on data across multiple trust domains.

GenAI systems introduce a fundamental architectural shift: system prompts, user inputs, retrieved documents (RAG), tool outputs, embeddings, and agent memory are aggregated into a single context window without internal access controls. This fusion of trust domains creates new categories of data exposure, integrity risk, identity sprawl, and governance failure that traditional data security frameworks were not designed to address