A claim of unauthorized model extraction tied to Claude shows how modern AI risk now includes not just prompt abuse, but the industrial harvesting of model behavior through ordinary API access.
Anthropic unmasks a sprawling network of Chinese AI labs siphoning millions of queries to clone its advanced language model.
Anthropic smaschera una vasta rete di laboratori cinesi di IA che sottrae milioni di query per clonare il suo modello linguistico avanzato.