Anthropic 称新防护措施将蒸馏窃取量降低 90%-99%
Anthropic 给 Claude 加了防蒸馏防护,声称砍掉 90% 以上抽取量,还点名了智谱和月之暗面,2 万个假账号这个数字挺猛的。
Anthropic 表示新部署的防护措施让竞争对手实验室能从 Claude 中蒸馏提取的信息量减少了 10 到 100 倍,即约 90% 到 99%。公司点名了中国的智谱(Zhipu)和月之暗面(Moonshot),并披露其中一个网络曾同时运营超过 2 万个虚假账号。该声明将直接影响 GLM 和 Kimi 在难以继续蒸馏情况下的模型迭代表现。
Anthropic says new safeguards have cut the distillation that rival labs can pull out of Claude by 10 to 100 times ("around 90 to 99 percent").
Anthropic names Chinese labs Zhipu and Moonshot, and says one network once ran more than 20,000 fake accounts at a time.
It remains to be seen whether GLM and Kimi will continue to release outstanding models when they can barely distill anymore.