Unverified50% confidenceCautionExact time
Policy engines rely on agents to truthfully report their behavior; if the underlying execution log is tampered with, the compliant behavior seen by the policy engine may be a false report.
1
Sources
50%
Confidence
Long-term
Relevance
10/11/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedPolicy engines verify compliance rather than execution itself; they can confirm that an agent was permitted to do something but cannot prove that the thing actually happened as recorded.76% similarUnverifiedModels claim to be rigorous and reliable but hallucinate and fabricate facts, claim to be friendly but exhibit sycophantic deference to users, and claim to be compliant but deviate from instructions on complex tasks.70% similarUnverified由于大语言模型存在提示词注入攻击风险和幻觉问题，权限过大的Agent可能在被恶意输入诱导或判断失误时对业务数据、外部系统造成不可逆破坏70% similarUnverifiedCurrent tool selection strategies rely heavily on the model's immediate judgment and lack systematic optimization; in enterprise scenarios where available tools can number in the tens to hundreds, models are prone to hallucination, repeated calls, or selecting the wrong tool.69% similarUnverifiedIn a LangGraph multi-agent system, passing complete conversation transcripts during agent handoffs wastes tokens and inference compute, and introduces misreading risks, because downstream agents need to rediscover which decisions have been confirmed, which tool results are authoritative, and which issues remain unresolved.69% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/998861API
curl https://kongchang.com/api/v1/knowledge/claims/998861MCP
get_claim(id=998861)