Free template
AI register template: list every AI tool you use
An Excel register with filled-in examples, a screening sheet that sets each tool's risk tier, a training log and an incident log. It also tells you which tools belong in your privacy policy from 10 December 2026.
Or read the full text free, below.
Version 1.0 · Updated · Reflects Australian law and guidance as at 30 October 2026 · Not legal advice
In short
An AI register is a list of every AI tool and use in a business, with an owner, a purpose, the data it touches, where that data goes and a risk level. It shows where AI is used, including AI built into software you already pay for, and it is the starting point for every other part of AI governance.
What is in the workbook
| Sheet | What it is for |
|---|---|
| Start here | How to use the workbook, in seven steps |
| AI register | One row per AI tool and use, with seven filled-in examples for a fictional 40-person Brisbane accounting practice |
| Screening | Eight questions, answered in order, that set each use's risk tier: Low, Medium, High or Prohibited |
| Training log | Who has accepted the AI use policy and completed AI training, and when the refresher is due |
| Incident log | AI errors, data leaks and complaints, the data breach assessment and what you did |
Dropdowns keep entries consistent, every heading on the register sheet has a note explaining what to enter, and status colours always sit next to a text label.
The columns, and why each one matters
| Column | Why it matters |
|---|---|
| AI tool or system, vendor, account type | Free and personal accounts give you the least control over your data. Recording the account type shows where staff are using the wrong one. |
| Owner | One named person answers for each tool: its settings, its users and its risks. |
| What we use it for, who uses it | The same tool can be low risk for drafting and high risk for screening applicants. One row per use keeps them apart. |
| Most sensitive data it touches | Decides which information rules apply, and whether the Privacy Act's automated decision rule can apply. |
| Where the data is stored or processed | Disclosing personal information to a provider overseas brings in APP 8. Unless an exception applies, you can be held responsible if that provider mishandles it. |
| Does the vendor train its AI on our data? | Check the contract, not just the settings page, and record when you checked. |
| Decisions about people, and whether they are significant | The two tests that, with the data column, decide whether a use belongs in your privacy policy. |
| Privacy policy: list as an automated decision? | Calculated for you. "Yes" means describe it in your privacy policy from 10 December 2026. |
| Records conversations; customers interact with it | Triggers the recording consent rule and the duty to tell people they are dealing with AI. |
| Risk tier, controls, limits | What keeps the use safe, in plain words. |
| Approved by and on, users trained, next review | The evidence that someone accountable looked at it. |
How to fill it in
- Ask every team what they use. Include free tools, browser extensions, AI notetakers, and AI features switched on inside software you already pay for: email, documents, CRM, accounting, HR and recruitment.
- Add one row per tool and use. Read the grey example rows first, then delete them.
- Screen each row with the Screening sheet. Stop at the first "Yes": that row gives the tier.
- Fix what the register shows you. Typical first fixes: move staff from personal to business accounts, switch off provider training, and pause any decision about people that nobody reviews.
- Review it every quarter, and whenever a vendor changes its terms, model or price.
How the register finds automated decisions
The Privacy Act rule that starts on 10 December 2026 applies when three things are true: a computer program makes, or does something substantially and directly related to making, a decision; the decision could reasonably be expected to significantly affect someone's rights or interests; and personal information about them is used.
The register asks those questions in three columns and calculates the answer, using the most sensitive data column for the third test. "Not sure" counts as "Yes" until you have checked, which follows the OAIC's advice to include a decision in your privacy policy when in doubt. Two of the example rows show results that surprise people:
- CV screening in recruitment software is flagged, even though a person reviews the shortlist. The rule covers programs that substantially help make a decision, not only fully automated ones.
- Automatic payment terms in practice management software are flagged, even though they are simple rules and not AI. The rule says "computer program", and the OAIC's guidance covers rule-based processes, with a spreadsheet formula among its worked examples.
The four risk tiers
| Tier | Typical uses | What it needs |
|---|---|---|
| Low | Drafting and summarising public or internal information, with a person checking the output | Pre-approved for tools on the approved list; add a register row |
| Medium | Personal or confidential information goes in; customers see the output; content is published or sent to clients | AI owner approval, vendor checks, settings confirmed, users trained |
| High | Decisions about people; sensitive information; recording or monitoring; AI agents acting on their own | Approval by the AI owner and a senior leader, a written impact assessment, testing and a plan for human oversight |
| Prohibited | Deepfakes of real people, covert recording, decisions about people with nobody involved, unapproved facial recognition | Not allowed |
The government's AI screening questions treat a "yes" to any of seven questions as a reason for more oversight. One question asks whether a system is general purpose, so nearly every use of ChatGPT or Copilot is flagged. The tiers above keep that scrutiny for the uses that need it.
What is in the register
General information, not legal advice. Have a lawyer review your final versions. Read the disclaimer
The Excel workbook has these columns, examples and questions, with dropdowns and a calculated privacy column.
The columns you fill in
| Column | What to enter |
|---|---|
| ID | A short reference for this row, used in the incident log and reports. |
| AI tool or system | The product name and, if it matters, the version or plan. For example: ChatGPT Business, Microsoft 365 Copilot, website chatbot. |
| Vendor or source | Who provides it. For AI built into other software, name that software's vendor. |
| Account type | Free and personal accounts usually give you the least control over your data. Business plans usually let you switch off training and manage users. |
| Owner (accountable person) | One named person who answers for this tool: its settings, its users and its risks. |
| Status | Proposed: being considered. Pilot: a small trial. In use: approved and running. Paused or Retired: no longer used (keep the row as a record). |
| What we use it for | The job it does, in plain words. Be specific: 'draft client emails from bullet points' is better than 'productivity'. |
| Who uses it | Teams or roles, and roughly how many people. |
| Most sensitive data it touches | Pick the most sensitive kind of information that goes into or comes out of the tool. Sensitive information under the Privacy Act covers racial or ethnic origin, political opinions or membership of a political association, religious or philosophical beliefs, trade union or professional association membership, sexual orientation or practices, criminal record, health and genetic information, and biometric information used to identify or verify people. |
| Kinds of personal information used | If personal information goes in, list the kinds in plain words (for example: names and contact details, payment history, CVs). Your privacy policy lists kinds of information, so this column feeds it. |
| Where the data is stored or processed | If overseas, the Privacy Act's cross-border rules (APP 8) can apply. Name the country in the notes. |
| Does the vendor train its AI on our data? | Check the contract and the admin settings. Record what you found and when. |
| Retention set to | How long the tool keeps prompts, files, recordings or outputs, as set in its settings or contract. Match your record-keeping rules (policy Schedule 1). |
| Vendor terms checked on | The date you last checked the vendor's contract and settings for training, retention and data location. |
| Does it make, or help make, decisions about people? | Yes if its output decides, scores, ranks, filters, recommends or prioritises something about an individual. A person reviewing the output does not, on its own, change the answer. Only extensive review (the person questions the output, uses other evidence and records why they depart from it) might. If you are not sure, answer "Not sure". |
| Could that decision significantly affect someone's rights or interests? | For example: whether someone gets a job interview, credit, payment terms, insurance, a service, a price, a tenancy or access to their account. Small effects, like the order of search results, are usually not significant. |
| Does a person make the final decision? | Yes if a person decides and the program only scores, ranks or recommends. No if the program's output takes effect with no person involved. Your privacy policy lists these two kinds separately. |
| Privacy policy: list as an automated decision? (calculated) | Yes when the tool uses personal information to make, or substantially help make, a decision that could reasonably be expected to significantly affect a person. 'Not sure' counts as Yes until you have checked. The answer also says which of the two lists in your privacy policy the decision belongs in. |
| Records or transcribes conversations? | AI notetakers, call recording and transcription. Recording laws differ by state: get consent from everyone before you record. |
| Customers or the public interact with it? | If yes, tell people they are dealing with AI, and give them a way to reach a person. |
| Risk tier | Set this with the Screening sheet. |
| Tier check (calculated) | Flags a row whose answers point to High risk under the Screening sheet but whose tier is lower. |
| Controls and rules for this tool | What keeps it safe. For example: business plan only, training switched off, no client data, a person checks every output before it is sent. |
| Known limits and banned uses | What the tool must not be used for, and what it does badly. For example: not for advice to clients; makes up references. |
| Approved by | The person who approved this use. Medium uses: the AI owner. High uses: the AI owner and a senior leader. |
| Approved on | The date of approval. |
| Users trained? | Have the people who use this tool completed the AI training and signed the policy? Record names in the Training log sheet. |
| Next review | When to review this row. Quarterly for High uses, at least yearly for the rest, and whenever the vendor changes its terms, model or price. |
| Notes | Anything else: the countries where data is held, what you checked and when, open questions for the vendor. |
The workbook also records the ID, status, known limits and banned uses, who approved each use and when, whether users are trained, the next review date and notes.
The worked examples
Seven rows for a fictional 40-person Brisbane accounting practice, from a low-risk drafting tool to automatic payment terms. The privacy policy answer is calculated in the workbook from the data, decision and significance columns.
ChatGPT Business
Risk tier: Medium · Privacy policy (APP 1.7): No
- Account type
- Business or enterprise plan
- What we use it for
- Draft and edit emails, proposals and blog posts from our notes; summarise public documents.
- Who uses it
- All staff (40)
- Most sensitive data it touches
- Confidential information
- Where the data is stored or processed
- Overseas (named in notes)
- Does the vendor train its AI on our data?
- No (confirmed in contract or settings)
- Does it make, or help make, decisions about people?
- No
- Could that decision significantly affect someone's rights or interests?
- No
- Controls and rules for this tool
- Business workspace only; personal accounts banned for work. Client information only where the client has given permission (for tax agents, Code item 6 can require it: see TPB(GS) 55/2026). No tax file numbers or bank details. A person checks every output before it leaves the firm.
- Notes
- Data processed in the US (record this in the privacy policy's overseas disclosure section). Training setting checked in admin console on 3 Nov 2026.
Microsoft 365 Copilot
Risk tier: Medium · Privacy policy (APP 1.7): No
- Account type
- Business or enterprise plan
- What we use it for
- Summarise long email threads and Teams chats; draft documents from files staff can already open.
- Who uses it
- Pilot group (8)
- Most sensitive data it touches
- Personal information
- Where the data is stored or processed
- Unknown: check with vendor
- Does the vendor train its AI on our data?
- Unknown: check with vendor
- Does it make, or help make, decisions about people?
- No
- Could that decision significantly affect someone's rights or interests?
- No
- Controls and rules for this tool
- Fix over-shared SharePoint and Teams folders before rollout: Copilot can surface anything a user can access. Pilot users only.
- Notes
- Confirm data location and training terms with our Microsoft partner before full rollout.
AI meeting notetaker
Risk tier: High · Privacy policy (APP 1.7): No
- Account type
- Business or enterprise plan
- What we use it for
- Record, transcribe and summarise client meetings.
- Who uses it
- Partners and managers (10)
- Most sensitive data it touches
- Personal information
- Where the data is stored or processed
- Overseas (named in notes)
- Does the vendor train its AI on our data?
- Opted out (setting turned off)
- Does it make, or help make, decisions about people?
- No
- Could that decision significantly affect someone's rights or interests?
- No
- Controls and rules for this tool
- Tell everyone at the start and get their consent; stop if anyone objects. Never auto-join meetings. Delete recordings after 90 days; keep the approved summary in the client file.
- Notes
- Recording laws differ by state. Consent script is in the AI use policy, Schedule 3.
Website chat assistant
Risk tier: Medium · Privacy policy (APP 1.7): No
- Account type
- Built for us (custom)
- What we use it for
- Answer common questions about our services and book appointments.
- Who uses it
- Public
- Most sensitive data it touches
- Personal information
- Where the data is stored or processed
- Australia
- Does the vendor train its AI on our data?
- No (confirmed in contract or settings)
- Does it make, or help make, decisions about people?
- No
- Could that decision significantly affect someone's rights or interests?
- No
- Controls and rules for this tool
- Says it is an AI assistant. Answers only from our approved FAQ. Hands over to a person on request. No pricing promises or tax advice.
- Notes
- If the assistant tells a customer something wrong or misleading, we can be liable under the Australian Consumer Law. Test answers every month.
CV screening in our recruitment software
Risk tier: High · Privacy policy (APP 1.7): Yes: describe in privacy policy
- Account type
- Built into software we already use
- What we use it for
- Rank job applicants against the job description.
- Who uses it
- HR (2)
- Most sensitive data it touches
- Personal information
- Where the data is stored or processed
- Overseas (named in notes)
- Does the vendor train its AI on our data?
- Unknown: check with vendor
- Does it make, or help make, decisions about people?
- Yes
- Could that decision significantly affect someone's rights or interests?
- Yes
- Controls and rules for this tool
- On hold until: bias testing, a person reviews every rejected applicant, privacy policy updated, applicants told AI is used.
- Notes
- Anti-discrimination laws apply to our decision, whatever tool helped make it. Tell applicants that AI helps screen applications.
Automatic payment terms in our practice management software
Risk tier: High · Privacy policy (APP 1.7): Yes: describe in privacy policy
- Account type
- Built into software we already use
- What we use it for
- Rules that put new clients on upfront payment or 14-day terms based on their payment history.
- Who uses it
- Finance (3)
- Most sensitive data it touches
- Personal information
- Where the data is stored or processed
- Australia
- Does the vendor train its AI on our data?
- No (confirmed in contract or settings)
- Does it make, or help make, decisions about people?
- Yes
- Could that decision significantly affect someone's rights or interests?
- Yes
- Controls and rules for this tool
- Approved as a High risk use because the rules decide with no person involved. Clients can ask for a review by a person. Rules tested for unfair results before launch and every six months. Reasons recorded for each review.
- Notes
- Not AI, but the Privacy Act rule covers any computer program, including simple rules. List it in the privacy policy.
AI writing in our design software
Risk tier: Low · Privacy policy (APP 1.7): No
- Account type
- Business or enterprise plan
- What we use it for
- Draft social media posts and image captions from public information about our services.
- Who uses it
- Marketing (2)
- Most sensitive data it touches
- Public information only
- Where the data is stored or processed
- Overseas (named in notes)
- Does the vendor train its AI on our data?
- Opted out (setting turned off)
- Does it make, or help make, decisions about people?
- No
- Could that decision significantly affect someone's rights or interests?
- No
- Controls and rules for this tool
- Public information only. A person checks every post before it is published.
- Notes
- Screening: question 9 (none of the above).
The screening questions
Answer in order and stop at the first “Yes”. If every answer is “No”, the tier is Low.
| No. | Question | If yes | What it means you must do |
|---|---|---|---|
| 1 | Would the use break the law or our policy's prohibited list? For example: deepfakes of real people, covert recording, monitoring staff without the notice the law requires, or facial recognition without approval. | Prohibited | Do not use it. Tell the AI owner. |
| 2 | Will it make, or substantially help make, a decision about a person that could significantly affect them (a job, credit, payment terms, insurance, a service, a price, a tenancy, discipline), including decisions it makes with no person involved? | High | Approval by the AI owner and a senior leader. Written impact and bias check. A person reviews outcomes, or for decisions the program makes on its own, people can ask a person to review them. Describe it in the privacy policy (APP 1.7) if the Privacy Act applies. We do not let a program decide recruitment, pay, discipline or termination on its own. |
| 3 | Will sensitive information go into it (for example health, genetic or biometric information, racial or ethnic origin, religion, political opinions, trade union membership, sexual orientation or criminal record), or information about children? | High | Owner approval. Usually consent is needed to collect sensitive information. Business plan only, with training switched off and data location confirmed. |
| 4 | Will it record or transcribe conversations, or monitor staff? | High | Consent from everyone recorded. Written notice to staff before any monitoring (14 days in NSW and the ACT). Check the state recording law. A notetaker tool with the policy's consent script can be approved once as a kind of use, not meeting by meeting. |
| 5 | Will it act on its own: send messages, change records, make payments or call other systems without a person approving each action? | High | Limits on what it can do. Logs. A person can pause or stop it at any time. Test before launch. |
| 6 | Will customers or the public deal with it directly, or see its output without a person checking it first? | Medium | Tell people it is AI. Give them a way to reach a person. Test answers regularly: you are responsible for what it says. |
| 7 | Will personal or confidential information go into it? | Medium | Approved business account only. Training switched off. Data location recorded. Least data needed. |
| 8 | Will its output be published, sent to clients or used to give advice? | Medium | A person checks every output for accuracy, copyright and tone before it goes out. |
| 9 | None of the above: public or internal information only, a person checks the output, and nothing leaves the business unchecked. | Low | Pre-approved if the tool is on the approved list. Add it to the register. |
Use this template
Download the editable Excel file, or the full kit with all five templates.
Questions
Frequently asked questions
More from the kit
The rest of the Australian AI Governance Kit
Word
AI policy template for Australian businesses
Free AI use policy template for Australian businesses, with staff rules, data limits, notetaker consent and privacy law built in. Read it, then download.
Read it free
Word
Automated decisions and your privacy policy: what to change by 10 December 2026
From 10 December 2026, privacy policies must explain significant automated decisions. Check if APP 1.7 applies to you and adapt our sample wording.
Read it free
AI governance checklist for Australian businesses
Plain-English AI governance checklist for Australian businesses, mapped to the six practices in the Guidance for AI Adoption, with the law behind each.
Read it free
Disclaimer
General information, not legal advice. These pages and templates are general information only. They are not legal advice and do not create a lawyer-client relationship. Osher Digital is an AI and automation consultancy, not a law firm.
Current as at 30 October 2026. They reflect Australian legislation and regulator guidance as at that date. Laws, guidance and vendor terms change, so check that the content is still current before you rely on it. Adapt the templates to your business, and have a qualified Australian lawyer review your final versions, especially if you work in a regulated industry or use AI or other software to make decisions about people.
Liability. To the extent permitted by law, Osher Digital is not liable for any loss or damage, including loss caused by negligence, arising from the use of or reliance on these pages or templates. Nothing in this disclaimer excludes, restricts or modifies any right or remedy you have under the Australian Consumer Law that cannot be excluded.
Sources and endorsement. The kit covers the six essential practices in the Australian Government's Guidance for AI Adoption (National AI Centre, October 2025). Links to government and other websites are for convenience; we do not control them. Osher Digital is not affiliated with or endorsed by the Australian Government, the National AI Centre or the Office of the Australian Information Commissioner.
Permitted use. You may copy, adapt and use the templates inside your own business or organisation. You may not sell them, or republish them as templates for others, without our written permission.
Want a second pair of eyes on your AI register?
Book a free 15-minute call. We look at your register, tell you which tools to fix first, and show you where AI can save your team time without adding risk.