Research

CamoDocs: A Poisoning Attack Against Retrieval-Augmented Language Models Using Camouflaged Documents

Page published

Publication date unknown · First observed: 17 Nov 2026

Coverage timeline

17 Nov 2026aclanthology.orgobserved

Single-source research — one report is available.

Why it matters

CamoDocs highlights the risk that attacker-controlled retrieval sources can compromise RAG answer integrity without modifying the underlying model.

CamoDocs presents research on poisoning retrieval-augmented language models using camouflaged documents. The available abstract describes attackers injecting malicious documents into public or user-editable sources to steer answers, but is truncated before details of the proposed method or evaluation.