Unverified50% confidenceCautionExact time
In multi-tenant RAG scenarios, if vector retrieval does not filter by tenant at query time, cleverly worded questions may retrieve documents belonging to other customers, resulting in cross-tenant data leakage.
1
Sources
50%
Confidence
Long-term
Relevance
10/11/2026
First Seen
Sources
Related Entities
Related Claims
Unverified向量数据库多租户隔离中，共享的Embedding模型可能导致语义层面的侧信道泄露66% similarUnverifiedDifferent phrasings of user queries will generate different vectors, and the semantic distance from document content will also vary, affecting RAG retrieval performance.65% similarUnverified大多数RAG（检索增强生成）实现只解决了'检索到什么'的问题，却忽视了检索到的内容是否可信、是否互相矛盾的更深层问题64% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/998094API
curl https://kongchang.com/api/v1/knowledge/claims/998094MCP
get_claim(id=998094)