Unverified50% confidenceFactExact time
RAG架构引入了间接提示注入攻击面：攻击者只需在RAG检索的数据源中植入恶意指令即可操纵模型
1
Sources
50%
Confidence
Long-term
Relevance
7/17/2026
First Seen
Sources
Related Claims
Unverified间接提示注入攻击中，攻击者只需在RAG会检索到的数据源中植入恶意指令即可操纵模型，传统WAF无法拦截83% similarVerifiedIndirect prompt injection conceals attack instructions within external documents, web pages, or database records retrieved by the model, and is activated when a RAG system introduces them into the context82% similarUnverifiedThe bottleneck of a RAG system often lies not in retrieval itself, but in how to tell the model via the prompt how to utilize the retrieved content.72% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/541069API
curl https://kongchang.com/api/v1/knowledge/claims/541069MCP
get_claim(id=541069)