Retrieval Poisoning

Retrieval poisoning is an attack that corrupts the external data sources an agent retrieves from — such as knowledge bases, vector stores, or document repositories — to manipulate the agent's behavior through its RAG pipeline.