Governed RAG is retrieval-augmented generation in which retrieval and context assembly are policy-controlled: chunks/elements/information are filtered, redacted or annotated by role, intent and sensitivity before reaching the model and every answer carries citations and lineage.
What is governed RAG?
Standard RAG retrieves whatever is most relevant, regardless of whether the requester is allowed to see it. Governed RAG inserts policy enforcement between retrieval and generation, so answers stay within each user's boundaries and remain auditable. It is the difference between a RAG demo and a RAG application that survives security review.