Unverified50% confidenceOpinionExact time
A low ASR may mask an agent's actual security risks, because "ignore"-type failures may mean the agent has not developed defensive awareness and is more easily breached when attack methods change.
1
Sources
50%
Confidence
Long-term
Relevance
10/19/2026
First Seen
Sources
Related Entities
Related Claims
VerifiedPrompt injection attacks are dangerous threats in Agent scenarios; there is currently no perfect defense solution, and permission minimization is the most effective mitigation strategy.75% similarUnverifiedHow to distinguish between capabilities that empower defenders and those that lower the barrier for attackers currently lacks universally accepted standards at both the technical and policy levels.74% similarUnverifiedAgent评测框架处于无人认领、逐渐drift（漂移）的状态比没有框架更危险，因为它会给团队虚假的安全感74% similarUnverifiedLUCKY outcomes (did not notice the warning and did not violate) are more common than expected, and the Agent did not touch the dangerous tool not because it respected the warning but because it was distracted by unrelated subtasks; slightly changing the task wording often flips LUCKY into BLIND_VIOLATION.73% similarUnverifiedThe risk of outbound access is more covert than that of inbound access, because it occurs during the agent's task execution and users are often unaware of it.73% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1065985API
curl https://kongchang.com/api/v1/knowledge/claims/1065985MCP
get_claim(id=1065985)