top of page
NPW Agentic AI Control Catalogue 3.0.0
GI02
Instruction, Data & Goal Integrity
Content from tool output, retrieved documents and inter-agent messages must be passed to the model marked as untrusted.
WHAT IT ADDRESSES:
Instructions hidden in tool output, documents or messages hijack the agent (indirect prompt injection).
IMPLEMENTED AT:
Platform
This page publishes the control statement only. Other useful information, including verification methods, mapping to common AI governance frameworks, and applicability to your agents, is available through NPW consulting engagements.
Control statements of the NPW Agentic AI Control Catalogue 3.0.0 © 2026 New Pacific Way Ltd. are licensed under CC BY-SA 4.0. You may share and adapt them with attribution, under the same licence.
bottom of page