Model distillation is a technique in which a smaller, faster AI model is trained to imitate a larger, more capable one. This hub explains how distillation works, why it matters for cost and efficiency, and the disputes it raises when done without permission.
A distillation attack is when someone systematically harvests a proprietary AI model's outputs — at massive scale — to train a clone of it without authorization. Here's how it works and why it's a growing concern.
Read more →Alibaba is banning employees from Anthropic's Claude Code effective July 10 after Anthropic revealed that accounts linked to Alibaba's Qwen lab ran 28.8 million unauthorized interactions with Claude in what the company called its largest ever distillation attack.
Read more →Knowledge distillation lets a small, fast AI model learn to think like a much larger one — and it now sits at the center of some of the biggest disputes in AI.
Read more →Anthropic told US senators that operators linked to Alibaba used nearly 25,000 fake accounts and 28.8 million exchanges to illicitly copy Claude's capabilities into Alibaba's Qwen model — calling it the largest known attack of its kind.
Read more →Anthropic told US senators that Alibaba-linked operators ran 25,000 fraudulent accounts and conducted 28.8 million exchanges with Claude — calling it the largest known AI distillation attack in history.
Read more →