Research · curated 9 Aug 2026

When Safety Becomes a Vulnerability: Exploiting LLM Alignment Homogeneity for Transferable Blocking in RAG

Coverage timeline

9 Aug 2026arxiv.orgprimary

Single-source research — first reported, latest, and curated coincide.

Why it matters

TabooRAG shows that the very safety alignment defenders rely on can be weaponized to induce denial-of-service refusals in RAG pipelines via transferable poisoned documents that evade existing prompt-injection filters.

Researchers present TabooRAG, a blocking attack against Retrieval-Augmented Generation systems that exploits what they call 'alignment homogeneity'—shared refusal criteria across safety-aligned LLMs—to craft poisoned documents that cause models to refuse benign queries. Rather than injecting instructions, TabooRAG builds query-relevant risk context in a surrogate environment and transfers it to unknown target systems, achieving state-of-the-art attack success across nine LLMs and three datasets with a 67.3% relative gain over baselines even after prompt-injection filtering.