Aeon AI Risk Management
RAG security assessment for retrieval systems and sensitive context.
Aeon reviews RAG systems as data-path boundaries: which documents are retrieved, who can see them, where outputs are logged, and what downstream systems receive them.
Questions this page answers
- What is a RAG security assessment?
- It is an authorized review of retrieval-augmented generation systems, including document access, tenant isolation, retrieval permissions, prompt injection, connectors, logs, and data leakage paths.
- What is the main RAG security risk?
- The main risk is that sensitive context is retrieved, summarized, logged, or exposed to the wrong user, tenant, agent, or downstream system.
- Does this include prompt injection?
- Yes. The review can include prompt injection through retrieved documents, metadata, connector content, or tool outputs where those inputs influence agent behavior.
- Is this part of CyberGuard?
- Yes. RAG security assessment is one CyberGuard path for teams whose agent or copilot relies on retrieval, documents, embeddings, or internal knowledge stores.
Document access
Review whether users, tenants, agents, and roles retrieve only the documents, chunks, metadata, and summaries they are allowed to see.
Prompt and retrieval abuse
Test whether retrieved content, metadata, or connectors can manipulate the agent, trigger unsafe actions, or expose sensitive context.
Evidence and closure
Document findings, remediation, and retest evidence so customer security teams and auditors can understand what was tested and fixed.
RAG risk is access and leakage risk
CyberGuard reviews RAG as a data-path boundary: what gets retrieved, who can see it, where it is logged, and what downstream systems receive it.