module three
AI systems and organizational AI governance.
Six sub-modules, fifty minutes. Assumes Module 2.
Funding narrative — Q2
v14
Program delivery — reporting period two
A. Point — confirm the allocation figure
Records/Funding
Locked for review — version history retained
what it covers
Writing an AI use policy.
What a system does with the data you give it, how open-weight, closed and self-hosted deployment differ, and which uses need a human in the loop. Six sub-modules, ending in two policy sections drafted for your own organization. Staff are already putting organizational information into these systems; the policy exists before that becomes a problem, or after.
01
Trace the submitted data
Follow submitted information through a vendor’s stated architecture and name every party with access to it.
02
Choose a deployment posture
Open weights, closed models or self-hosting, with the break-even worked between owned and rented compute.
03
Assign the risk tiers
Where a system may inform a decision, where it may draft but not decide, and where it should not be present.
04
Draft the policy sections
The prohibited data classes and the tool approval gate, written for the organization rather than copied from anyone else.
the sub-modules
Sub-modules.
Each one stands alone: a cold open on a real document, five to eight minutes of teaching, one activity on your own material, and five graded items.
Model data handling
Context, training data and retained logs, and every party in the chain with access to them.
Licensing and deployment
Open weights, closed models and self-hosting, with the break-even arithmetic between owned and rented compute.
continued
Sub-modules, continued.
The last three sub-modules move from how the systems behave to the policy a Nation adopts to govern them.
Allocations FY2026
Sheet 1
Programme
Budget
Committed
Housing
412,000
388,140
Lands & Resources
260,500
191,220
Health services
305,000
305,000
Education
178,400
96,850
Total
1,155,900
981,210
Figures drawn from live records — no re-entry
Language and cultural data
The highest-risk data class, and the stewardship models that keep control and benefit with the source.
Risk assessment and human review
Where a system may inform a decision, where it may draft, and where it should not be present.
Drafting the AI use policy
Scope, permitted and prohibited uses, approval gates, member disclosure, incident handling and review cycle.
The module artifact
A prohibited data classes list and a tool approval gate, drafted for the learner’s own organization.
the take away
What the learner keeps.
Every sub-module ends in a downloadable written to be usable by someone who has not taken the course.
Prohibited classes
The classes named, the prohibition stated, and the exception process set out where one exists.
module artifact
Approval gate
The step a new AI tool has to pass before staff may use it with organizational data.
take away
Module credential
Sub-modules are individually completable and individually badged, and all ten modules make the credential.
badged
next step
Take this module on its own.
Modules are bought singly and the introduction is free. All ten make the credential.