Unverified50% confidenceOpinionExact time
工具调用使攻击面从让模型说错话升级为让模型做错事，危害从信息误导上升为实质性的数据访问和系统篡改
1
Sources
50%
Confidence
Long-term
Relevance
7/9/2026
First Seen
Sources
Related Claims
VerifiedData poisoning attacks involve mixing malicious samples into training datasets, causing the model to exhibit incorrect behavior toward specific inputs after deployment.79% similarUnverified下游用户对模型的二次微调可能意外破坏原有安全特性74% similarUnverified系统会试探性地应用补丁，若失败则将错误信息反馈给模型进行修正，构成闭环纠错机制74% similarUnverified如果LoRA权重被逆向分析，攻击者可能从中推断出训练所用的私有数据内容，即记忆提取问题73% similarUnverified数据投毒攻击按攻击目标可分为可用性攻击和目标攻击，前者旨在降低模型整体性能，后者只在特定触发条件下改变模型行为71% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/386356API
curl https://kongchang.com/api/v1/knowledge/claims/386356MCP
get_claim(id=386356)