Unverified50% confidenceOpinionExact time
OWASP LLM Top 10 leans toward risk classification and compliance benchmarking, while MITRE ATLAS leans toward attack behavior modeling.
1
Sources
50%
Confidence
Long-term
Relevance
10/14/2026
First Seen
Sources
Related Entities
Related Claims
Partially VerifiedOWASP lists prompt injection as the top risk in the LLM Top 10 (LLM01).73% similarUnverifiedReference model-free attacks achieve, in most scenarios, results statistically comparable to those of attacks that rely on adversarial knowledge.65% 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65% similarUnverifiedMVT通过将设备日志与已知恶意指标数据库IOC比对进行检测，只能检测已知威胁，对全新手段或已清除痕迹的变种有效性大打折扣63% 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61% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1030522API
curl https://kongchang.com/api/v1/knowledge/claims/1030522MCP
get_claim(id=1030522)