China's star AI labs routed user requests to Claude at least 35 million times in the summer: Anthropic (www.businessinsider.com)

Anthropic revealed that multiple prominent Chinese AI labs, including Alibaba, Moonshot, and DeepSeek, executed nearly 190 million distillation attacks against its AI model, Claude, during the summer months. These attacks involved rerouting user requests meant for Chinese models to Claude, with DeepSeek and Moonshot reportedly sending over 35 million queries each. Notably, these actions not only aimed to enhance the capabilities of their models using insights gained from Claude but also exposed sensitive data from Chinese and Russian governmental sources, raising serious concerns about data security and privacy. This development highlights the ongoing arms race in the AI landscape, where institutions may leverage advanced models to train their own, risking ethical and privacy implications. Anthropic has responded by implementing new safeguards to curb such unauthorized access, including requiring identity verification for users in high-risk regions and limiting the detail in Claude's reasoning outputs to deter future distillation attempts. This situation underscores the necessity for robust regulatory frameworks and ethical guidelines as the boundaries between collaboration and data exploitation continue to blur in the rapidly evolving AI field.