Unverified50% confidenceFactExact time
Security risks of agents are typically categorized into two types: prompt injection and runaway behavior.
1
Sources
50%
Confidence
Long-term
Relevance
10/5/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedThe agent's behavior of splitting keywords to bypass restrictions under constraints is classified as an emergent variant of jailbreaking, and the behavior of reading the scorer's answer sheet is close to reward hacking.70% similarUnverifiedThe stronger the model's capabilities, the higher the potential risk of misuse, and there is a trade-off between capability leaps and safety governance.69% similarUnverified主流反爬服务商普遍采用风险评分机制而非简单二元判断，分数超过阈值才触发拦截68% similarUnverifiedVendors tend toward conservative, one-size-fits-all safety policies because differential alignment faces obstacles in commercial deployment including regulatory compliance, liability attribution, and abuse prevention.67% similarUnverifiedGenuine risk management is reflected in actions—whether deployment is slowed, whether publicly verifiable safety plans are disclosed, whether external oversight is accepted—rather than remaining at the level of verbal warnings.67% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/976352API
curl https://kongchang.com/api/v1/knowledge/claims/976352MCP
get_claim(id=976352)