Part 7/14:
A crucial aspect of this saga involves foreign espionage and the theft of AI models. Anthropic admits that Chinese entities are actively stealing their models via illicit distillation attacks. In these sophisticated maneuvers, Chinese labs create fake accounts, interact with Claude extensively, and reverse engineer the models to extract logic and capabilities. It's estimated that over 16 million exchanges were used to siphon off proprietary AI understanding, fueling China's own military and surveillance applications.