美国AI标准机构：Kimi K3网络安全能力落后于美国前沿模型，安全防护仍允许漏洞利用开发

Odaily Planet Daily News: The US Center for AI Standards and Innovation has released an evaluation stating that Kimi K3, developed by Moonshot AI, significantly lags behind leading US frontier large language models in cybersecurity capabilities. Additionally, the evaluation pointed out that Kimi K3's security protection mechanisms still allow the generation of content that can be used for vulnerability exploitation development. (Cointelegraph)