Unverified50% confidenceBenchmarkExact time
The study shows that even when explicitly told to be vigilant against threats, the human miss rate remains as high as approximately 33%.
1
Sources
50%
Confidence
Long-term
Relevance
8/7/2026
First Seen
Sources
AI代理指令审批失效：人类漏放三分之一威胁的深层原因与对策
Related Claims
Unverified25%的风险降幅数字仍需在方法论严格的复制研究中验证，应被视为值得进一步验证的信号而非最终定论68% similarUnverifiedIn a 2024 research paper, Anthropic quantified the phenomenon of model over-refusal, finding that safety-trained models had false positive refusal rates as high as 15–30% in certain scenarios63% similarUnverified对抗样本攻击由Szegedy等人于2013年首次系统性揭示，只需叠加人眼不可见的微小扰动（幅度仅像素值1%~2%）即可让检测模型准确率从80%跌至接近随机水平60% similarUnverifiedAcademic research indicates that the impact of survivorship bias on backtesting results may be as high as 1–3 percentage points of return bias per year.60% similarUnverified研究表明当虚假告警率超过30%时，团队对告警的响应速度会显著下降59% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/702860API
curl https://kongchang.com/api/v1/knowledge/claims/702860MCP
get_claim(id=702860)