Secure AI Atlas mark Secure AI Atlas SECURITY & GOVERNANCE
EN / ES

RAG Knowledge Assistant

Mapped through AI Capability → Exposure Surface → Control Layer → Governance Layer.

AI Capability

  • Document ingestion
  • Semantic retrieval
  • Answer synthesis

Exposure Surface

  • Internal documents
  • Vector store
  • Access control boundaries
  • Tenant data

Risk

prompt injectionsensitive data disclosurevector embedding weaknessessystem prompt leakage

Control

data classification before aiinput output validation agentsprompt output logging

Framework

Open questions

  • OQ-EVIDENCE-ADEQUACY

This mapping is deterministic: it comes from relationships already present in the ATLAS catalogue. No model was used to infer it.