Testing a Prompt injection Attack Against an Enterprise AI Agent

This blog shows how Darktrace’s behavioral approach to email security was able to detect a prompt injection attack that contained no malware malicious links, or sender reputation indicators.