Event ·Multi-source

Kimi K3在网络安全漏洞利用上大幅落后美国前沿模型，蒸馏或许是原因

英国AI安全研究所与美国AI标准与创新中心测试了Moonshot AI的Kimi K3在进攻性网络任务上的表现。Kimi K3在ExploitBench上得分32%，而领先美国模型得分76%。其安全防护未能阻止漏洞利用开发或模拟攻击。其通用基准测试高分与网络安全弱项之间的差距也符合Moonshot AI蒸馏Anthropic模型的指控。

Current read

Kimi K3在网络安全测试中只拿了32分，远低于美国模型的76%，而且安全机制基本失效，传闻是因为用了蒸馏技术。

11 sources67° AI heatLast updated 2026/7/24 09:48:32

Evidence

11 sources

Primary sourceDecoder

2026/7/24 09:48:32View source

Related source 1The Rundown AI

2026/7/22 15:32:12View source

Related source 2@koltregaskes

2026/7/23 09:53:04View source

Related source 3techcrunch

2026/7/23 11:00:00View source

Related source 4IT之家

2026/7/22 11:51:15View source

Related source 5Paul Couvert

2026/7/22 16:23:07View source

Related source 6shao__meng

2026/7/23 01:22:07View source

Related source 7歸藏(guizang.ai)

2026/7/23 03:15:57View source

Related source 8Nathan Lambert: Interconnects

2026/7/22 14:09:04View source

Related source 9Simon Willison

2026/7/23 01:29:00View source

Related source 10Together AI

2026/7/23 21:38:44View source

Conflict check

Dedup data only shows that these sources discuss the same event, not that they agree. There is no structured supporting or refuting evidence yet, so no conflict is inferred automatically.

Open the story