>Anthropic 公布反蒸馏防御措施
Start here
An Anthropic research article states that early evidence shows its new distillation defense method reduces the amount of data obtained by distillers by approximately 90% to 99%, and the number of reasoning tokens extracted on the platform has dropped to 1/10 to 1/100 of the previous level. The defense falls into three categories: using preserved thinking to prevent new API accounts from editing Claude's prior context, and making reasoning records visible only to the organization that produced them; using a two-stage classifier to intercept extraction attempts, with a false positive rate below 0.1%; and tightening limits on and banning accounts that repeatedly attempt, with charges applied to intercepted requests starting September 24. Anthropic's official X account subsequently stated that over the past few months it has made clear progress in reducing the effectiveness of distillation attacks against Claude, and attached a link to the research article.
Generated by AI from reports · Updated 3 hr ago
Progress timeline
- X @AnthropicAIAnthropic says it has made progress in reducing the effectiveness of distillation attacks against Claude
Anthropic's official X account said that over the past few months it has made notable progress in reducing the effectiveness of distillation attacks against Claude, and attached a link to the research article anthropic.com/research/distillation-defenses, stating that the article explains the specific approach.
- Anthropic ResearchSelectedAnthropic strengthens defenses against distillation attacks
Anthropic says early evidence shows its new distillation defenses reduced the amount of data obtained by distillers by approximately 90% to 99%, with extracted reasoning tokens on the platform dropping to 1/10 to 1/100 of previous levels. The defenses fall into three classes: using preserved thinking to prevent new API accounts from editing Claude's prior context and making reasoning records visible only to the organization that produced them; using a two-stage classifier to intercept extraction attempts with a false positive rate below 0.1%; and tightening rate limits and banning accounts that repeatedly attempt extraction, with charges applied to intercepted requests starting September 24. Anthropic says that on SWE-bench Verified, a model trained with summarized reasoning scored 71.8%, while one trained with full reasoning scored 80.4%. Anthropic also says Zhipu distilled the cyber capabilities of a U.S. model while developing GLM-5.3, and CAISSI assessed it as the open-weight model with the strongest cyber capabilities to date. These measures require some developers to update their integrations.
Related discussion
Attention trend
New sources per day
- Oct 12New1 sources (1 outlets,0 accounts)
- Oct 13New2 sources (1 outlets,1 accounts)