Anthropic 谈蒸馏攻击防御

AI 导读

Anthropic 称过去几个月在降低针对 Claude 的蒸馏攻击有效性方面取得显著进展，并发布文章说明具体做法。Nathan Lambert 批评 Anthropic 此前为寻求监管帮助大造声势，并质疑其一边强调从自家有漏洞的 API 学习有害行为的安全风险、一边继续托管该 API。

正文 · AI 翻译

Nathan Lambert· @natolambert · X·2026-10-13 21:38· 1 小时前AI 评分36

AI 导读

我觉得很奇怪，Anthropic 在做这件事之前，为了寻求监管帮助造了那么大的声势。希望现在能重新开启对话。更奇怪的是，他们还强调从自家有漏洞的 API 学习有害行为的安全风险。如果 API 不安全，就别托管它。

正文

It’s bizarre to me how much noise anthropic made asking for regulatory help before doing this. Hopefully can be a conversation reset now.

Was doubly bizarre bc they emphasized the security risk of learning harmful behaviors from their leaky api. If API is unsafe don’t host it.

引用Anthropic@AnthropicAI

Over the last few months, we’ve made significant progress in reducing the effectiveness of distillation attacks against Claude. In this post, we explain how: https://www.anthropic.com/research/distillation-defenses

在 X 查看被引用的帖子

来源：Nathan Lambert · x.com

来源：RadarAI · aihot.news