Ga naar de hoofdinhoud

Prompt-Injection Battlecards

Recall practical responses to common prompt-injection incident objections.

ai-incident-response5 min read

It's just text They say: 'It's just a line in a PDF, not an exploit.' Use this when teams underplay indirect prompt injection. Your line It is text in the channel the model treats as instructions. The risk is whether that text can reach tools, data, or users without a boundary. Judging the artifact instead of the path it can influence. It reframes prompt injection as a trust-boundary problem. Prompt says no They say: 'The system prompt already tells it not to follow malicious instructions.' Use this when prompt hardening is presented as containment. Good instruction helps, but containment needs…

Read the full lesson

Sign up free: one personalized lesson every day, matched to your role and goals.

← Back to library

Neem contact op