artifacts/intake-archive/20260622__continuity-office-intake/011-ai-mandates-boundaries

Rendered from markdown source. Open raw source on GitHub.

--- catalog: "Free Training Catalog" training_id: "011" title: "AI Mandates & Boundaries" subtitle: "Consent continuity for automated systems" track: "AI & Automation Continuity" estimated_time: "20–30 minutes" audience:

• Executives
• Product
• IT / Security
• Compliance
• AI teams

learning_outcomes:

• Define clear mandates for AI systems
• Prevent silent scope expansion
• Preserve consent through automation

prerequisites: "Training 001–010 recommended" level: "Intermediate" license: "Free / Open Training" version: "1.0" last_updated: "2025-12-18" ---

AI Mandates & Boundaries

Consent continuity for automated systems

Core stance

AI systems should never be allowed to answer the question:

“Since I can do this, may I?”

Mandates and boundaries ensure AI only acts within explicitly granted authority.

What an AI mandate is

An AI mandate is a plain-language statement of:

• What the system is allowed to do
• For whom
• Using what inputs
• For what purpose

If this cannot be stated clearly, the AI system is over-scoped.

What boundaries prevent

Boundaries stop:

• Silent reuse of data
• Expansion into adjacent decisions
• “Temporary” pilots becoming permanent authorities

The two-sentence rule

Every AI system must have:

1. A mandate sentence (what it may do)
2. A boundary sentence (what it must not do)

Example:

This system may summarize customer support tickets for internal quality review. It must not generate customer-facing responses or be used for performance evaluation.

Exercises

• Write mandate + boundary sentences for one AI tool
• Identify one current AI use with no explicit boundary
• Add a revocation or review trigger

Suggested next step

Publish one AI mandate publicly inside your organization.