Safety Alignment as Attack Surface

RAGOCR competes with RAG / Shared entity: RAG / Same source domain / What happened next / Tension

RAGOCR renders retrieved documents as images and beats naive RAG by 15% on one-eighth the input tokens.

Linked by a graph relationship (RAGOCR competes with RAG); both cover RAG; reported by the same outlet (arxiv.org).

Shared entity: RAG / Same source domain / Shared topic / What happened next / Tension

TopoGuard catches 21x more RAG split-knowledge attacks than LlamaGuard-2-8B at sub-millisecond latency.

Both cover RAG; reported by the same outlet (arxiv.org); overlapping topics (attack, benign, document).

Shared entity: RAG / Same source domain / What happened next / Tension / Downstream implication

Test your RAG under corpus mutation, not against a static snapshot.

Both cover RAG; reported by the same outlet (arxiv.org); picks up the RAG thread on 2026-07-30.

Shared entity: RAG / Same source domain / Shared topic / What happened next

Red-team your RAG with deliberately poisoned retrieval.

Both cover RAG; reported by the same outlet (arxiv.org); overlapping topics (inject, knowledge).

Shared entity: RAG / Same source domain / What happened next / Tension

GraphRAG detection plans survive full IOC rotation where naive RAG collapses to 29%.

Both cover RAG; reported by the same outlet (arxiv.org); picks up the RAG thread on 2026-08-14.

Perplexity-based RAG poisoning detection is actively misleading, because poisoned answers score lower perplexity.

Both cover RAG; reported by the same outlet (arxiv.org); picks up the RAG thread on 2026-08-10.

Per-chunk RAG verification is worse than no filtering at all

Both cover RAG; reported by the same outlet (arxiv.org); picks up the RAG thread on 2026-08-04.

Semantic chunking loses to simple chunking in new RAG test

Both cover RAG; reported by the same outlet (arxiv.org); picks up the RAG thread on 2026-07-21.