Purpose
Controls for retrieval trust, source contamination, prompt injection, sensitive data leakage and citation reliability.
FRAMEWORK
RAG Security Assurance Layer
Controls for retrieval trust, source contamination, prompt injection, sensitive data leakage and citation reliability.
framework routeevidence-oriented modelhuman review required
Evidence artifacts
Use the model to collect decision records, assumptions, risk boundaries, review checkpoints and operational evidence before relying on the output.
Use limits
This framework supports structured review. It does not replace accountable expert review, certification, regulatory approval or operational sign-off.
RELATED TOOLS
Instruments that operationalize this model.
Open a tool to turn the framework into an evidence checklist, readiness score or decision artifact.
Agentic AI and RAG Security
RAG Contamination Threat Model
Maps contamination paths from source material through retrieval into model output and downstream action.
open instrument → Agentic AI and RAG SecurityAI Vendor Assurance Questionnaire Builder
Generates buyer or vendor assurance questions for model APIs, copilots, RAG platforms and agentic workflow vendors.
open instrument → Agentic AI and RAG SecurityAgentic AI Risk/Readiness Safety-Case Engine
Builds separated risk, readiness and confidence safety-case artifacts for tool-using AI agents.
open instrument → Agentic AI and RAG SecurityAgentic Workflow Rail
Builds a swimlane control map for user, agent, policy, memory, retrieval, tools, human approval, audit and rollback.
open instrument → Agentic AI and RAG SecurityFrontier Agent Evaluation Runner / Export Contract
Defines executable evaluation tasks, datasets, scorers, logs and JSON export contract for frontier/agentic systems.
open instrument → Agentic AI and RAG SecurityPrompt-Injection Sandbox
Classifies suspicious instruction patterns in test text and maps them to defensive containment recommendations.
open instrument →