Prompt Injection Isn’t Magic. It’s an AI Following the Wrong Instructions
中文摘要
提示词注入并非魔法，而是AI执行了错误指令。实验表明间接注入极易触发，且对AI代理构成威胁。
English Summary
Prompt injection happens when AI follows wrong instructions. Experiments reveal that indirect injection is easily triggered and dangerous for AI agents.
Original Excerpt
A simple document experiment shows why indirect prompt injection is so easy to trigger — and why AI agents make it dangerous Continue reading on Medium »