Session 5 of 6. An applied session.
SAFE RESPONSIBLE

Build your personal AI policy. From the Victorian public sector framework. From your organisation. From the shape of your role.

FacilitatorDr Tiffany Gray
CohortIPAA Victoria·20 participants
Time3:00 to 3:55·55 min
DateWednesday 13 May 2026
People lead·AI follows
Session 5 · Introduction

Safety is what unlocks the work.

Generative AI is becoming embedded in everyday work across the Victorian public sector. The Department of Premier and Cabinet issues the Administrative Guideline. The Office of the Victorian Information Commissioner sets the privacy rules. Public Record Office Victoria sets the records rules. Australia’s national AI Ethics Principles set the broader expectations.

You operate inside all of these. Your role has its own context. Your organisation has its own policy. The goal of this session is not to memorise the rules. It is to make them yours. To build a personal AI policy you can attach to every conversation you have with AI from tomorrow onwards.

This is not about being cautious for caution’s sake. It is about being safe in a way that lets you move forward. Safe and responsible use is what unlocks AI as a working partner. Without it you hesitate. With it you proceed with confidence.

Session 5 · Purpose

Why we are doing this.

Many people read the rules once and forget them. Or they assume the rules are too restrictive and stop trying. Both miss the point. The rules are the conditions for confident use. This session helps you internalise them by applying them to your actual role and your actual work.

Purpose 01
Translate the framework into actions

Turn the Victorian public sector AI guidance into specific actions you actually take.

Purpose 02
Identify what goes in and what stays out

Get clear on what information is appropriate to share with AI in your role and what is not.

Purpose 03
Build a verification routine

Design a verification routine that fits how you actually work. Not a checklist you ignore.

Purpose 04
Compile a personal AI policy

Walk out with a personal AI policy you can save and reuse for every AI session from tomorrow.

Session 5 · Learning Objectives

By the end of this module.

Four things you will be able to do.

Objective 01
Apply the Victorian framework

Apply the Victorian AI guidance to your specific role and organisational context.

Objective 02
Identify safe information handling

Distinguish safe and unsafe information handling for AI use in your everyday work.

Objective 03
Define your escalation paths

Name who you escalate to when AI use crosses into uncertain territory.

Objective 04
Build your personal AI policy

Compile a personal AI policy that can be used as standing context for every AI session.

The Victorian framework · Where the rules come from

Four documents that set the rules.

Before we build your personal policy. Know where the rules come from. The Victorian Public Sector has clear, principles-based guidance. It applies to your work. These four documents are the foundation.

Document 01
Administrative Guideline

Endorsed by the Victorian Secretaries Board in September 2024. Issued by the Secretary of DPC. Sets minimum requirements and safeguards for GenAI across the VPS. Principles-based and broad.

Document 02
OVIC Privacy Guidance

From the Office of the Victorian Information Commissioner. Sets privacy and information security obligations for enterprise GenAI tools. Aligned to the Victorian Protective Data Security Standards.

Document 03
PROV Records Guidance

From Public Record Office Victoria. Tells you what AI-assisted work becomes an official record and how to keep it. Audit trail expectations are explicit here.

Document 04
National AI Ethics Principles

Victoria signed up in June 2024. Five principles: wellbeing, human-centred values, fairness, privacy and security, reliability and safety. The foundation under everything else.

The five guardrails · Click each to reveal

Five guardrails. From the framework into practice.

The Administrative Guideline turns into five practical rules you can hold in your head. The same five rules apply across the VPS regardless of which agency you sit in. Click each number to read the rule.

01
Click to reveal
Guardrail 01
Approved tools first

Use agency-approved Copilot in your M365 tenancy wherever the work involves organisational material. Personal-licence tools are for public material only.

02
Click to reveal
Guardrail 02
Public info only outside

If you are not in your agency-approved tool, only publicly-available information goes in. No internal documents. No personal information. No commercial-in-confidence.

03
Click to reveal
Guardrail 03
People decide

AI never makes or substitutes decisions affecting specific individuals. Not performance assessments. Not selection. Not eligibility. The delegate decides. AI assists.

04
Click to reveal
Guardrail 04
Verify before you ship

Check AI outputs against reliable sources before they enter official work. Confidently-wrong is the danger. Plausible does not mean correct. Verification is a habit not a step.

05
Click to reveal
Guardrail 05
Audit trail

When AI-assisted content becomes an official record, capture the prompt, the output and the verification. PROV expects it. Your organisation expects it. So should you.

Before you start · Organise your work with AI

Name your chat.

A small habit. Big payoff. Every conversation you have with AI gets stored. The ones with names are the ones you can find again. Set the name before you do anything else.

For this session, name your chat
Session 5. My Personal AI Policy
Open Copilot. Start a new chat. Set the name. Then move to the start prompt.
Setting context · The start prompt

Start the conversation.

Paste this prompt into your new chat. It tells AI what session you are in, what you are about to build and the three steps we will work through.

The start prompt
// I’m starting session 5 of the GEN Ai AT WORK Training Program. This session is called Safe and Responsible Use. I’m a [ROLE TITLE] at [ORGANISATION NAME] [URL] in the Victorian public sector. I’m going to use you to help me build my own Personal AI Policy based on the Victorian public sector AI framework, my organisation’s policy and the specifics of my role. We will work through three steps. First. I will share my organisation’s AI policy and ask what I need to be aware of in my role. Second. We will work through real scenarios where I’m not sure whether AI use is appropriate. Third. We will compile my Personal AI Policy as a document I can save and attach to every future AI session. Tell me you understand and then we will begin.
Step 01 of 3 · Read your policy through your role

Read your policy through your role.

The prompt (after uploading your organisation’s AI policy)
// I’ve uploaded my organisation’s AI policy. Read it carefully. // Based on the policy and my role as [ROLE TITLE], tell me the top five things I personally need to be aware of when using AI in my work. // For each one, explain why it matters specifically to my role and what behaviour it requires from me. // Use plain English. Australian spelling. Be specific to my role. Not generic. // Ask me one clarifying question if you need more context about my role before answering.

What you are doing

You are taking a general policy document and asking AI to filter it through the specifics of your role. The same policy means different things to a procurement manager, a frontline service officer or a data analyst. You are getting AI to do the translation.

Why this matters

Policies are written for everyone. Which means they often feel relevant to no-one. Asking AI to read your policy through the lens of your specific work makes the rules tangible. You get the five things that apply to you, not the fifty that apply to someone else.

Step 02 of 3 · Work through real scenarios

Work through real scenarios.

The prompt
// Now I want to work through real scenarios from my work where I’m not sure whether AI use is appropriate. // Ask me to describe 3 to 5 situations I’ve encountered. One at a time. // For each one, ask one or two clarifying questions to understand the context. // Then give me a verdict using this exact format: Question: [the scenario in one sentence] Verdict: [Yes / No / Stop and check] Reasoning: [why, in 2 to 3 sentences, referencing the policy where relevant] Better path: [if no or stop and check, what to do instead] // Ask me for one scenario at a time. Wait for my answer before moving to the next.

What you are doing

You are stress-testing the policy against your actual work. Edge cases are where policies get tested. The scenarios that nag at you in the back of your mind are the ones that need a verdict before you face them at speed.

Why this matters

You build judgement through deliberate practice. Working through real scenarios with AI as a thinking partner trains you to spot the pattern. Next time a similar scenario lands on your desk you will already know which guardrail applies.

Step 03 of 3 · Compile your personal AI policy

Compile your personal AI policy.

The prompt
// Based on everything we have discussed so far. My role. My organisation’s policy. The things I need to be aware of. The scenarios we worked through. Build my Personal AI Policy. // Structure it as a document I can save and attach to every AI session from tomorrow. Use the following sections: 1. My role and context (one paragraph) 2. The information I can use AI with (clear lines, specific to my role) 3. The information I cannot use AI with (clear lines, specific to my role) 4. My verification routine (what I check before using AI output in official work) 5. My escalation paths (who I go to when in doubt. Privacy, legal, records, IT, my manager) 6. My voice rules (Australian English, plain language, any banned phrases specific to my role) 7. My decision boundary (what decisions I never let AI make or recommend for me) // Use Australian English. Plain language. Active voice. Make it specific to my role. Not generic. // When you are done ask me if I want to refine any section.

What you are doing

You are pulling together everything into a single document. Not a long policy that sits unread. A short, role-specific reference you can paste at the start of every AI conversation. The Victorian framework becomes yours.

Why this matters

A policy you carry is a policy that works. By compiling your own version you commit to the rules. You also build a reusable artefact. When you start a new AI thread tomorrow you paste your policy. AI knows your guardrails before you ask anything else.

Wrapping up · Same thread

Capture your policy.

Go back to the same AI thread you have been working in. Paste this prompt. Your AI now knows your role, your organisation’s policy implications, your scenario verdicts and your draft personal AI policy. It will help you finalise.

The wrap-up prompt
// We’ve just built my Personal AI Policy together. You now know my role, my organisation’s policy implications, the scenarios I worked through and the policy we have drafted. Before you finalise anything ask me three things. What stood out for me as I worked through this. What surprised me about my own AI use. What is the one rule I am most committed to following from tomorrow. Ask me the questions 1 at a time and wait for my answer before moving to the next question. Then ask whether I’d like the final version of my Personal AI Policy as a take-away document. If I say yes give it to me in a clean structured format I can save.
Now · 10 minutes · Groups of 4

Share what you have built.

The question

The rule you did not realise you needed. The escalation path you will set up tomorrow. The one thing you will never let AI do for you.

01
The rule you did not realise you needed

Two minutes each. The thing you put in your policy that surprised you. The line you almost crossed without knowing.

02
The escalation path you will set up

Privacy. Legal. Records. IT. Your manager. Which conversation will you have first to set up that path so it is ready when you need it.

03
The one thing AI will never do for you

Your decision boundary. The one call you will always make as a human. Say it out loud so the room hears it.

Time10 minutes
Groups5 groups of 4
Bring backOne rule. One escalation path. One boundary.
Wrapping up · What you have done in 55 minutes

What you have done.

Not a compliance exercise. A working personal policy that fits the Victorian framework and your role. Something you carry forward into every AI conversation from tomorrow onwards.

Done 01
Read your policy through your role

Five things from your organisation’s policy that personally apply to you. Not the fifty that apply to someone else.

Done 02
Worked through real scenarios

Used AI to give you a verdict on the edge cases that have been nagging you. Yes. No. Stop and check. Reasoning attached.

Done 03
Compiled your personal AI policy

Role context. Information lines. Verification routine. Escalation paths. Voice rules. Decision boundary. All in one document.

Done 04
Named your escalation paths

Privacy. Legal. Records. IT. Your manager. You know who to go to when in doubt before you are in doubt.

Done 05
Drew your decision boundary

The one thing AI will never do for you. The line you hold as a human. Said out loud.

Done 06
Built a reusable artefact

A document you save. Attach to every future AI conversation. Update as your role evolves. Carry forward.