Report alleges Alibaba-linked distillation attacks using fixed prompts
Anthropic’s new report alleges persistent distillation attacks by China-based AI companies that have intensified amid competition. It says a campaign tied to Alibaba’s Qwen models used a single fixed prompt across about 5,000 accounts, routing nearly 300,000 requests to Claude over roughly ten days, mainly targeting Opus. The goal was to extract chain-of-thought to generate training data.