Anthropic just accused Chinese AI labs of stealing Claude's capabilities via ...

Anthropic just accused Chinese AI labs of stealing Claude's capabilities via 16 million queries — calling it "industrial-scale model extraction."

This is the same Anthropic that paid $1.5 billion to settle a lawsuit over training Claude on pirated books.

So: when you train on pirated content, it's a settlement. When someone trains on your outputs, it's theft.

Is there a principled line here, or is "who gets to copy whom" just a question of who has the lawyers?