社区讨论LinuxDo
Anthropic 高度评价 GLM-5.3 的网络安全能力,同时批评其安全护栏过于脆弱
Anthropic 高度评价 GLM-5.3 的网络安全能力,同时批评其安全护栏过于脆弱 anthropic.com – 29 Sep 26 GLM-5.3 and the spread of advanced cyber capabilities GLM-5.3 can autonomously build end-to-end cyber exploits, but unlike other frontier models, it was released without meaningful safegua…
内容摘要
Anthropic 高度评价 GLM-5.3 的网络安全能力,同时批评其安全护栏过于脆弱
anthropic.com – 29 Sep 26
GLM-5.3 and the spread of advanced cyber capabilities
GLM-5.3 can autonomously build end-to-end cyber exploits, but unlike other frontier models, it was released without meaningful safeguards to limit misuse.
总结:
GLM5.3 和 GLM5.3Flash 都非常强
GLM5.3 的网络能力已经是中国模型的 Mythos 级别
GLM 的安全护栏极其脆弱,简单的骗,或者直接修改模型权重,就可以几乎完全删掉护栏(他们预估,熟练团队这样改权重可能只需要花费 1200 美元)
于是,Anthropic 呼吁各国政府对强大模型进行独立审查
68 个帖子 - 36 位参与者
阅读完整话题
via LINUX DO - 热门话题 (author: lueluelue)
相关链接:
https://www.anthropic.com/research/glm-5-3-and-the-spread-of-advanced-cyber-capabilities
https://linux.do/t/topic/2968325