·12 min read

What Should an AI Acceptable Use Policy Say? Free Template for Australian Businesses

An AI acceptable use policy should name the tools and accounts staff may use, the information that must never go into them, when a person checks the output, who approves new tools and how it meets the Privacy Act. A free 11-section template with Australian clauses, in Word with no sign-up.

An AI acceptable use policy should say five things: which AI tools staff may use and with which accounts, what information must never go into them, when a person has to check AI output before it is used, who approves a new tool, and how AI use meets your Privacy Act obligations. For an Australian business it should also cover overseas disclosure of personal information, the automated-decision disclosure that starts on 10 December 2026, and, for anyone holding health information in Victoria, the Health Records Act 2001. Two pages that staff read beat twenty that nobody opens.

Below is a complete template in 11 sections that you can copy, followed by how to enforce it in Microsoft 365. It is also available as a Word document, with no email or sign-up. It is general information, not legal advice; if you handle sensitive information at scale, have your lawyer read your final version.

Why a short policy is enough to start

The OAIC's guidance on privacy and the use of commercially available AI products (published 21 October 2024, updated 17 January 2025) makes three points that a policy has to answer:

  • Privacy obligations apply to any personal information put into an AI system, and to AI output that contains personal information.
  • Businesses should establish policies and procedures for their use of AI systems, and update their privacy policies with clear information about how they use AI.
  • As a matter of best practice, organisations should not enter personal information, and particularly sensitive information, into publicly available generative AI tools.

Staff are already using AI whether or not there is a policy. A short policy that answers "can I use this, with this information, for this job?" is more useful than a long one that only says "be careful". Our AI adoption and governance guide covers the wider framework; this page is the policy itself.

The government template, and what this one adds

The National AI Centre publishes an AI policy guide and template as a Word document (published 22 April 2026). It covers principles for responsible use, expected behaviours, roles and responsibilities, approval points, and monitoring and review. It is a good structure, and deliberately general.

The template below is narrower and more prescriptive. It assumes a business of a few to a few hundred people that runs on Microsoft 365, and it fills in the parts the general template leaves blank: an approved-tools table, a list of banned information, Australian privacy clauses, and rules for automations that act on their own.

The template: AI acceptable use policy

Replace everything in [square brackets]. Delete any clause that does not apply to you, and keep the numbering so staff can refer to it.

1. Purpose and scope

This policy sets out how [Business name] staff may use artificial intelligence (AI) tools for work. It applies to all employees, contractors and volunteers, on any device, whenever they use an AI tool for [Business name] work or with [Business name] information.

The policy owner is [name and role]. Questions go to the policy owner.

2. Definitions

  • AI tool: any software that generates text, images, audio, code or decisions from a prompt or from data, including chat assistants, writing and meeting assistants, AI features inside other software, and automations that call an AI model.
  • Approved tool: an AI tool listed in section 3, used with the account listed there.
  • Personal information: information or an opinion about an identified or reasonably identifiable individual, as defined in the Privacy Act 1988.
  • Sensitive information: sensitive information as defined in the Privacy Act 1988, such as health, genetic and biometric information, racial or ethnic origin, political opinions, religious beliefs, sexual orientation, criminal record, and union membership.
  • Confidential information: anything [Business name] or a client would not publish, including financial records, contracts, pricing, passwords, and client files.

3. Approved tools and accounts

Only the tools below may be used for [Business name] work, and only with the account shown. Using an approved tool with a personal account counts as using an unapproved tool.

ToolAccountApproved forNot approved for
Microsoft 365 Copilot ChatYour [Business name] work accountDrafting, summarising and research with internal informationSensitive information unless section 7 is met
Microsoft 365 Copilot (licensed users only)Your [Business name] work accountWorking with documents, email and Teams you already have access toFiles you were not meant to see; report them under section 10
[Other tool, for example Claude Team or ChatGPT Business][Business account administered by the business][Specific uses][Specific exclusions]
Any free or personal AI accountNot applicablePublic information only, such as rewording a published web pagePersonal, sensitive or confidential information

Microsoft says Copilot and Copilot Chat used with a work account are covered by enterprise data protection: prompts and responses are covered by the same contractual terms as your email and files, and are not used to train foundation models. Check the equivalent terms for any other tool before you add it to this table.

4. Information you must not put into AI tools

Unless a tool in section 3 is expressly approved for it, do not enter, paste, upload or connect:

  • personal information about clients, patients, staff or job applicants;
  • sensitive information of any kind, including health information;
  • passwords, multi-factor codes, API keys or other credentials;
  • bank, card or tax file numbers;
  • information a client has given us under a confidentiality obligation;
  • unreleased financial results, pricing or tender responses;
  • [add anything specific to your industry].

If you are unsure, remove the names and identifying details first, or ask the policy owner.

5. Human review

AI tools make mistakes and can state invented facts with confidence. Before AI output is used, a person must check it. In particular:

  • anything sent to a client, published or filed with a government body is read in full by the person sending it;
  • figures, dates, legal or technical statements and quotes are checked against the source;
  • code is reviewed and tested before it runs on [Business name] systems;
  • the person who sends or uses the output is responsible for it, not the tool.

6. Being open about AI use

  • Tell a client when AI has substantially produced work they are paying for, if they ask or if your engagement terms require it.
  • Any chatbot or AI assistant that customers interact with must be clearly identified as AI.
  • Do not use AI to impersonate a real person's voice, image or writing without their written consent.

7. Privacy obligations

  • Use for the original purpose. Personal information may only be put into an AI tool for the purpose it was collected for, or a related purpose the person would reasonably expect (APP 6). If in doubt, ask the policy owner.
  • Overseas disclosure. Many AI services process data outside Australia. Before personal information goes to a tool that may disclose it overseas, the policy owner must confirm that the arrangement meets APP 8.
  • Security. Approved tools must be used with work accounts protected by multi-factor authentication (APP 11).
  • Automated decisions. [If the business is covered by the Privacy Act.] From 10 December 2026, if a computer program uses personal information to make, or to do something substantially and directly related to making, a decision that could reasonably be expected to significantly affect someone's rights or interests, our privacy policy must say so. Any such use must be added to the AI register (section 8) so the privacy policy stays accurate.
  • Health information in Victoria. [Delete if not relevant.] Health information we hold is also covered by the Health Records Act 2001 (Vic). Under Health Privacy Principle 9 it may only be transferred to someone outside Victoria in the circumstances that principle allows, so health information must not go into any AI tool that stores or processes it outside Victoria unless the policy owner has approved it.

8. Requesting a new AI tool or use

To use a tool or a use that is not in section 3, send the policy owner:

  • what the tool is and what you want to use it for;
  • what information it would receive;
  • where the vendor stores and processes data, and whether data is used to train its models;
  • who in the business would use it, and who can see what it produces.

The policy owner records each approved tool and use in the AI register: [location of the register].

9. Automations and AI agents

An automation or agent that takes actions, such as sending email, updating records, moving money or making a decision about a person, must:

  • have a named owner who is responsible for it;
  • be approved and recorded in the AI register before it goes live;
  • run under a business account, not a personal one;
  • keep a log of what it did;
  • have a person approve any step where a mistake would be costly or would significantly affect a customer, employee or applicant.

10. Reporting problems

Report to the policy owner straight away if:

  • personal, sensitive or confidential information was put into a tool that is not approved for it;
  • an AI tool showed you information you should not have access to;
  • AI output caused, or nearly caused, a mistake that reached a client or the public.

Reporting a mistake quickly will not be treated as a breach of this policy. The policy owner decides whether the incident is an eligible data breach under the Notifiable Data Breaches scheme.

11. Responsibilities and review

  • Staff: follow this policy and ask when unsure.
  • Managers: make sure their team knows the approved tools and section 4.
  • Policy owner: keep section 3 and the AI register current, decide on requests, and review this policy every [six or twelve] months or when the law or our tools change.

Breaches of this policy are handled under [name of your disciplinary or conduct policy].

Approved by [name], [role], on [date]. Next review: [date].

Download the template as a Word document (no sign-up).

The Australian clauses, explained

APP 6 and APP 8: what goes in, and where it goes

The OAIC's guidance says that if personal information is put into an AI system, APP 6 limits its use to the purpose it was collected for, unless the person consents or would reasonably expect the secondary use. APP 8 applies when personal information is disclosed to someone overseas. Because many AI services process prompts outside Australia, section 7 sends both questions to one person rather than leaving them to each staff member.

APP 1.7 from 10 December 2026

From 10 December 2026, APP entities must describe in their privacy policy any computer program that uses personal information to make, or to do something substantially and directly related to making, decisions that could reasonably be expected to significantly affect individuals' rights or interests. An AI policy that requires every such use to be recorded in a register is the simplest way to keep the privacy policy accurate. Our guide to the Privacy Act automated decision-making rule explains who it applies to, which automations count and what to write.

Health Records Act 2001 (Vic) and HPP 9

The Victorian Department of Health says the Health Records Act applies regardless of the size of the business: there is no small business exemption. It covers health service providers such as GP clinics, psychologists and allied health practitioners, and any other organisation that holds health information, including employers holding staff health records. Health Privacy Principle 9 allows health information to be transferred to someone outside Victoria only in listed circumstances, such as where the recipient is bound by substantially similar principles, the person consents, or the organisation has taken reasonable steps to make sure the information will not be handled inconsistently with the HPPs. For a Melbourne clinic, that is a question to answer before any AI tool, transcription service or offshore backup receives patient information.

How to enforce it in Microsoft 365

A policy staff can ignore is a suggestion. In a Microsoft 365 business, the controls that back it up are documented in Microsoft's guide to managing generative AI apps and its shadow AI deployment model. Which ones you have depends on your licences; our Microsoft 365 licensing guide compares the plans.

Policy sectionControlWhat it does
3. Approved toolsMicrosoft Defender for Cloud AppsLists the generative AI apps in use, with risk scores. Marking an app as unsanctioned blocks it on devices onboarded to Defender for Endpoint, or warns users instead.
3. AccountsMicrosoft Entra ID sign-inStaff use Copilot Chat and other approved tools with their work identity, so enterprise terms and your access policies apply.
4. Banned informationMicrosoft Purview Endpoint DLPBlocks pasting or uploading sensitive information to AI websites on managed devices. Microsoft recommends running it in simulation mode first.
4. Banned informationSensitivity labels with encryptionOther AI apps cannot decrypt content encrypted by Purview, so a labelled file pasted or uploaded elsewhere cannot be processed.
8 and 10. Register and reportingPurview DSPM for AI, activity explorer and audit logsShows user interactions with generative AI sites and DLP matches, which gives the policy owner something to review.

Before giving anyone a Copilot licence, check what it will be able to see: Copilot works with the permissions users already have, so overshared SharePoint sites become searchable. Our Copilot readiness checklist covers that clean-up, and the cybersecurity page covers the wider controls.

Common mistakes

  • Banning AI outright. Staff move to personal accounts on their phones, where you have no visibility at all.
  • Approving a tool but not the account. A business plan and a free personal account for the same product usually come with different data terms.
  • No owner. If nobody keeps the approved-tools list current, it is out of date within months.
  • Forgetting automations. A flow that screens applicants or puts accounts on credit hold is AI use too, and from 10 December 2026 it may need to be in your privacy policy.
  • Writing it and not telling anyone. Walk the team through sections 3 and 4 in a staff meeting and put the policy where people will find it.

About Nublock

Nublock Digital, the trading name of Nublock Group Pty Ltd, is a Melbourne-based business that provides AI consulting and automation alongside managed IT and Microsoft cloud services, and is a Microsoft CSP partner. For AI work, data stays in Australia, client data is not used for training and nothing is routed through Nublock accounts. Each AI engagement is quoted as a fixed price. To talk about AI in your business, get in touch for a 30-minute call.

Sources

All sources checked on 16 November 2026.

Does a small business in Australia need an AI policy?

If anyone in the business uses AI tools for work, yes. The Privacy Act applies to personal information put into an AI tool and to output that contains it, and the OAIC recommends that organisations establish policies and procedures for their use of AI systems. A two-page policy that names the approved tools, the banned information and who approves new tools is enough to start.

Is there a government AI policy template for Australian businesses?

Yes. The National AI Centre publishes an AI policy guide and template as a Word document on ai.gov.au (published 22 April 2026). It covers principles, expected behaviours, roles, approval points and review. It is general, so most businesses still need to add their own approved tools, banned data and privacy clauses.

Can staff use personal ChatGPT or other free AI accounts for work?

That is your decision, but the policy should answer it. The OAIC recommends, as a matter of best practice, that organisations do not enter personal information, and particularly sensitive information, into publicly available generative AI tools. Most businesses that allow AI therefore approve business accounts signed in with a work identity and treat personal accounts as not approved for work information.

Who should approve new AI tools in a small business?

Name one person, usually the owner, operations manager or whoever is responsible for privacy, and give them a short checklist: what the tool is for, what information it would receive, where the vendor stores and processes data, whether data is used for training, and who can see the outputs. Record each approval in an AI register so the approved-tools list stays current.

How do you enforce an AI policy in Microsoft 365?

With the controls Microsoft documents for managing generative AI apps: Microsoft Defender for Cloud Apps to discover AI apps and mark unapproved ones as unsanctioned, which blocks them on devices onboarded to Defender for Endpoint; Microsoft Purview Endpoint DLP to block pasting or uploading sensitive information to AI websites; sensitivity labels with encryption; and Entra ID sign-in so approved tools are used with work accounts. Which of these you have depends on your Microsoft 365 plan.