Lesson 4 of 12 · 5-minute lesson

Create and apply an AI policy

Create or review one controlled organisation policy and apply it where relevant.

For: Organisation administrators, executives and policy owners

This lesson uses National AI Centre guidance: Create an AI policy, AI policy guide and template.

The situation

Staff need clear expectations for approved use, human responsibility, information handling and incidents.

The problem

Informal rules and competing documents are difficult to apply or review consistently.

The choice

Maya creates one main policy with an owner, known version, adoption status and review trigger.

What you will learn

By the end of this lesson

  • Choose a supported policy starting point.
  • Create practical organisation-wide rules.
  • Adopt and apply one controlled version.

The pivotal moments

What changes—and why it matters

Moment 1

Choose one policy path

What the person does

Maya starts in the app or reviews existing material.

What the workspace records

Both paths lead to one main policy record.

Notice

Reviewing a document does not create a second policy system.

Why it matters

The organisation avoids competing policy records.

North Star policy library and supported policy starting paths
Supported starting paths lead to one main policy.

Moment 2

Write practical rules

What the person does

Maya covers acceptable use, accountability, information, incidents, monitoring and review.

What the workspace records

The workspace saves practical policy statements and governance responsibilities.

Notice

Policy language must tell people what to do.

Why it matters

Staff can apply expectations consistently.

Organisation policy record in North Star
Practical rules are controlled in one organisation record.

Moment 3

Publish a controlled version

What the person does

Maya confirms the owner, status, effective date and next review.

What the workspace records

Use cases refer to the adopted version and where it applies.

Notice

A policy applying to a use case does not prove that its safeguards work.

Why it matters

Teams use a known policy version during assessment.

Policy version and adoption state
The policy has visible ownership, state and review timing.

Official guidance used in this lesson

Check the published guidance

National AI Centre

Create an AI policy

Explains how an organisation can establish practical AI rules.

Last checked 30 September 2026

The complication

The difficult part stays visible

A policy can state the expected control, but it cannot demonstrate that people or systems perform it in practice.

The outcome

The organisation has one adopted, reviewable policy that use cases can reference.

Check your understanding

Does a relevant policy prove the use case is safe?

Show the answer

No. It sets expectations. The use case still needs evidence that its safeguards work.

Key points

Remember these three things

  • Maintain one main policy.
  • Control versions, ownership and review.
  • Test safeguards separately from policy use.

What happens next

Review the six Essential AI practices at the depth that fits the organisation.