Anthropic said on Monday that new measures designed to stop rivals from copying its Claude AI model capabilities have reduced the amount of data stolen in so-called "distillation attacks" by about 90% to 99%. According to Sina Finance, the company said safeguards aimed at limiting access to Claude's reasoning capabilities have sharply reduced extraction attempts, with the number of reasoning tokens extracted through distillation now down to between one-hundredth and one-tenth of the previous level. Anthropic said it has made it harder to repeatedly probe Claude's protected reasoning process through proxy networks, blocking data extraction attempts and hindering large-scale model distillation activity. The company also said its security system is roughly as robust as the system used to prevent Claude from assisting in the manufacture of biological weapons, and that its false positive rate is below 0.1%.