Unverified50% confidenceFactExact time
当恶意方占比接近或超过 1/3 时，鲁棒聚合算法理论上的安全保证将不复存在
1
Sources
50%
Confidence
Long-term
Relevance
8/19/2026
First Seen
Sources
抗投毒概念锚定：防御AI数据污染的新思路
Related Claims
Unverified安全对齐本质是在模型概率分布上施加偏置，而非从根本上消除有害能力73% similarUnverified纯粹基于软件的AI安全护栏存在被越狱攻击绕过的风险，对于触及关键能力阈值的模型延缓部署比仅依赖护栏更审慎66% similarUnverifiedThreshold-based defenses face the threat of borderline stealthy poisoning attacks, where an attacker can keep each sample just at the edge of the tolerance threshold and erode concepts through cumulative effect.66% similarUnverified身份验证与权限控制同时崩溃在安全工程中被称为'双重失效模式'，其整体风险远高于单一失效的线性叠加66% similarUnverified漏洞数量激增并不必然意味着模型更不安全，可能是审查强度提升带来的统计偏差65% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/772958API
curl https://kongchang.com/api/v1/knowledge/claims/772958MCP
get_claim(id=772958)