Sort The Control Layer
Choose the right control layer for an AI incident corrective action.
Place each corrective action into its primary control layer. Data control Model/prompt control Tool control Output control Workflow control Quarantine uploaded documents that contain instruction-like text before indexing. Revise the system prompt to refuse requests to reveal hidden instructions. Remove directory lookup permission from the assistant until approval logging is added. Block responses containing live API-key patterns before they reach users. Require human approval for AI-generated refunds above $500. Exclude obsolete policy versions from the retrieval index. Add a refusal instruction for unsupported medical triage recommendations. Route all AI-generated legal exception language to counsel review. Require scoped OAuth permissions for…
Sign up free — one personalized lesson every day, matched to your role and goals.
Already have an account? Sign in