Anthropic Cuts AI Distillation Attacks
Anthropic reports new security measures have reduced data harvesting from distillation attacks by up to 99 percent. The safeguards restrict access to Claude reasoning tokens to block rivals from copying advanced capabilities.
Xurve View
Insights:
Xurve View
Anthropic deployed new safeguards on Monday afternoon that cut data harvested in AI distillation attacks by 90% to 99%. The measures arrive amid intensifying technological competition between the United States and China over frontier models. The defense prevents rivals from cheaply copying advanced reasoning capabilities developed by industry leaders like Alphabet Inc. Class A.
