Unverified50% confidenceFactExact time
The security risk of influencing a model's output behavior by poisoning external documents it reads is known as 'Indirect Prompt Injection.'
1
Sources
50%
Confidence
Long-term
Relevance
8/6/2026
First Seen
Sources
TIME向AI爬虫投喂专属网页内嵌广告：媒体变现新策略
Related Claims
Unverified下游用户对模型的二次微调可能意外破坏原有安全特性64% similarVerifiedData poisoning attacks involve mixing malicious samples into training datasets, causing the model to exhibit incorrect behavior toward specific inputs after deployment.64% similarUnverified新接入的文档或知识库更新可能引入提示注入或数据投毒的新入口61% similarUnverified任何要求把文本喂进模型或管道的邀请都应视为潜在风险，切勿将来源不明数据接入生产系统或本地敏感环境61% similarUnverified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59% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/697199API
curl https://kongchang.com/api/v1/knowledge/claims/697199MCP
get_claim(id=697199)