Skip to content
← Back to Glossary
AI Security Checked by Vezraa

RAG Poisoning

Injecting malicious content into the knowledge base of a RAG system.

RAG (Retrieval-Augmented Generation) poisoning targets the external knowledge sources an LLM retrieves from. Attackers inject malicious documents that, when retrieved, manipulate the LLM's response. Sanitize and validate all documents in your RAG pipeline.

Related Terms

See if your app is vulnerable

Vezraa checks for this security issues. Scan in 25 seconds.

RAG Poisoning — Security Glossary | Vezraa