Aller au contenu principal

Map prompt risk before you optimize

Apply the NIST AI RMF govern-map-measure-manage loop to a prompt-based workflow.

ai-prompt-security5 min read

The move: run the risk loop before the prompt rewrite. The NIST AI RMF uses four functions that translate well to prompt security: Govern, Map, Measure, and Manage. Govern asks who owns the workflow and what risk is acceptable. Map asks what data, people, systems, tools, and harms are in scope. Measure asks how you will test for failure. Manage asks what control changes, monitoring, and escalation rules reduce the risk enough to operate. Govern Name the owner, reviewer, data steward, and escalation path. A prompt with no owner becomes nobody's control. Map List inputs, retrieved sources, model-visible context, outputs,…

Read the full lesson

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

← Back to library

Nous contacter