Anthropic Blocks Up to 99% of Claude Reasoning Token Thefts in Crackdown on AI Distillation
Anthropic has successfully reduced the unauthorized extraction of Claude's internal reasoning tokens by up to 99 percent, targeting industrial-scale distillation attacks linked to Chinese firms Zhipu and Moonshot. The company banned tens of thousands of accounts in the process, marking a major escalation in protecting proprietary AI models from intellectual property theft and competitive cloning.
Source: The Decoder