Unverified50% confidenceFactExact time
OWASP LLM Top 10中的'过度代理'（LLM06: Excessive Agency）条目专门针对智能体权限过度问题提供缓解指南
1
Sources
50%
Confidence
Long-term
Relevance
7/14/2026
First Seen
Sources
JadePuffer全自动AI勒索攻击实战复盘：智能体武器化时代来临
Related Claims
Unverified过度公开检测技术细节可能为恶意方提供规避蓝图59% similarUnverifiedExcessively frequent human intervention eliminates the efficiency advantages of automation, while an overly lenient approval scope fails to effectively prevent and control risks — this is the trade-off in HITL granularity design59% similarUnverified安全研究者将系统行动能力远超其安全判断能力的潜在风险称为能力越界（Capability Overhang）58% similarUnverifiedThe limit of a maximum of 9 concurrent sub-agents per round may be related to Anthropic's API rate limits and resource scheduling policies.58% similarVerifiedExcessive safety fine-tuning can cause models to over-refuse, undermining their practical utility.57% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/509489API
curl https://kongchang.com/api/v1/knowledge/claims/509489MCP
get_claim(id=509489)