Moonshot 在新 Kimi 模型中移除网络攻击相关训练数据
Moonshot curbs hacking ability in new Kimi model as fears grow over AI-powered cyberattacks
Why we picked it
Moonshot 新的 Kimi 模型主动砍掉了黑客相关训练数据，但保留了编程能力，是国产模型在安全上少见的做法。
Moonshot 表示在新 Kimi 模型训练中删除了语料库和强化学习环境中与网络攻击相关的数据，以限制模型的黑客能力。此举针对的是 AI 驱动网络攻击风险上升的担忧。同时公司强调保留了一般性编程能力，不影响正常开发用途。
Original · Rappler: Technology
Moonshot curbs hacking ability in new Kimi model as fears grow over AI-powered cyberattacks
Moonshot says it removed cyber-related data from the training corpus and reinforcement-learning environments used to develop the model, while retaining general coding skills
