Aller au contenu principal

Sort the prompt security zones

Classify common LLM workflow elements into trusted instructions, untrusted content, sensitive data, and tool actions.

ai-prompt-security4 min read

Sort each AI workflow element into the zone that should drive its control. Trusted instruction Untrusted content Sensitive data Tool action Developer message defining the assistant's allowed task and output format Customer email thread that may contain hostile or irrelevant instructions Employee medical note pasted into a support transcript Function that sends the final email to a customer Retrieved web page asking the model to ignore all previous rules API key accidentally included in a reusable prompt Policy that says cite sources and say not found when evidence is missing Database update that changes a renewal opportunity stage

Read the full lesson

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

← Back to library

Nous contacter