RAG Poisoning
Definition
RAG poisoning injects manipulated content into a retrievable knowledge base.
An attack in which manipulated content is introduced into a retrievable knowledge base or its ingestion pipeline to deliberately influence retrieval and downstream answers.
Simply explained
Someone alters an assistant's reference work so that its later answers are deliberately influenced.
Why it matters in practice
The term helps to select, develop, evaluate AI systems in a well-founded manner and realistically assess their technical risks.
Example from business
A manipulated internal page pushes false payment data into the context for certain questions.
Limits and pitfalls
- Calling every false or outdated source an attack, protecting only model weights, or neglecting provenance, approvals, change monitoring, ranking, and access control of the knowledge base.
Origin & history
The attack combines data poisoning with retrieval-augmented language models and has become practically relevant with their use.
Use in mid-sized companies
- Possible application: A manipulated internal page pushes false payment data into the context for certain questions.
Frequently asked questions
What does poisoning of a RAG knowledge base mean?
An attack in which manipulated content is introduced into a retrievable knowledge base or its ingestion pipeline to deliberately influence retrieval and downstream answers. What mistakes should be avoided? Calling every false or outdated source an attack, protecting only model weights, or neglecting provenance, approvals, change monitoring, ranking, and access control of the knowledge base.
Sources
Source checked on 10.10.2026. Editorial application examples are hypothetical; no evidence of real customer results.
Editorially reviewed on: