48% of security teams report blind spots around prompt injection chains or tool-chaining abuse in AI-native applications.

Some prompt injection incidents led to over $100,000 in real losses without requiring any code to be written.