The Decoder· Manuel Uth·· 3 d ago
Anthropic says it cut Claude reasoning token theft by up to 99% in fight against industrial-scale distillation
Anthropic says it cut Claude reasoning token theft by up to 99% in fight against industrial-scale distillation
AI summary
Anthropic says it has cut the extraction of Claude's internal reasoning tokens by 90 to 99 percent. The company links the large-scale distillation attacks to Chinese companies Zhipu and Moonshot, and has already banned tens of thousands of accounts.
Selection record
Threshold 76Media and individualsFirst 74Second 71
Not admittedSum of both 145 < twice the threshold 152
- Source tier
- Media and individuals; this tier's threshold is 76
- Pre-filter
- passed：Anthropic防御Claude模型蒸馏攻击，涉及AI模型能力窃取
A model scores each item twice, independently, against one written standard, out of 100. An item is admitted only when the two scores add up to twice the threshold. The threshold is set per source tier.
Source: The Decoder · the-decoder.com