AI vendor Anthropic says a China-backed threat group used the agentic capabilities in its Claude AI model to automate as much as 90% of the operations in a info-stealing campaign that presages how ...