For Heads-of · Practitioner

Vector and embedding weaknesses

Weaknesses in how vectors and embeddings are generated, stored, or retrieved in a retrieval-augmented generation pipeline let an attacker poison or exfiltrate the knowledge base.

medium
llm
rag
embeddings
owasp-llm

How it happens

A RAG pipeline's vector store lacks the access controls, input validation, or tenant isolation applied to the rest of the application, because it is treated as an implementation detail rather than a data store in its own right.

Why it matters

An attacker who can write to or query the vector store can poison what the model retrieves as ground truth, or extract embedded content that was never meant to be directly queryable.

Mitigating controls

The controls that address this risk, ranked by effectiveness.

primary

Vector store and embedding access controls

Access control, input validation, and tenant isolation applied to a RAG pipeline's vector store, the same as any other production data store.

Framework and clause references

Framework	Clause	Title
OWASP Top 10 for LLM Applications	LLM09	Vector and Embedding Weaknesses

Related resources

The external sources behind this risk, from the Resources library.

OWASP GenAI Security Project · framework

OWASP Top 10 for LLM Applications

Back to the risk library