Unverified50% confidenceFactExact time
Control evaluation is a systematic method used to quantify whether a safety protocol can remain safe under adversarial attack; the typical process assumes that learning at deployment time or an external attacker will cause some perturbation, the red team iteratively attacks, and if the protocol can still block dangerous actions it is considered to have a sufficient safety margin.
1
Sources
50%
Confidence
Long-term
Relevance
10/12/2026
First Seen
Sources
Related Entities
Related Claims
Unverified任何具备联网和代码执行能力的前沿模型在大规模部署前都应经过独立严格的安全评估，特别是针对网络攻击、供应链投毒等高危场景的专项测试76% similarUnverifiedCybersecurity evaluations are typically conducted in controlled sandbox environments for the purpose of testing a model's capabilities and tendencies in offensive and defensive scenarios.76% similarUnverifiedPermission checks must be based on normalized identifiers that are consistent with the actual operation; otherwise, they may be exploited by attackers.74% similarUnverifiedAgent security should be quantitatively evaluated using three core metrics: attack success rate, over-refusal rate, and attack type classification accuracy.73% similarUnverifiedSecurity reviews involve high-risk decisions, and agent conclusions should undergo human review, especially before the system has matured.72% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1012104API
curl https://kongchang.com/api/v1/knowledge/claims/1012104MCP
get_claim(id=1012104)