Anthropic accuses Alibaba of extracting Claude via 25,000 fake accounts

Anthropic has told US senators that operators linked to Alibaba used roughly 25,000 fraudulent accounts to pull about 28.8 million responses from Claude between April 22 and June 5 — what it calls the largest campaign yet to illicitly extract Claude's capabilities. Such 'distillation attacks' use a stronger model's answers to train a weaker rival, sidestepping export controls on model weights.