For AI privacy in Australia, the main law a business has to think about is the Privacy Act 1988: if the Act covers your business, its 13 Australian Privacy Principles apply to personal information you put into AI tools and to what those tools produce about people. The change to plan for starts on 10 December 2026, when privacy policies must explain certain decisions that computer programs make using personal information.
This is general information taken from the Act and its 2024 amendments on the Federal Register of Legislation. It is not legal advice. Check the current text, and speak to a privacy lawyer about your own situation.
Does the Privacy Act cover your business?
Start here, because the answer changes how much of the rest applies to you. The Privacy Act 1988 has a group of sections on small business: section 6D (small business and small business operators), section 6DA (how annual turnover is worked out), section 6E (small business operators treated as organisations) and section 6EA (small business operators choosing to be treated as organisations).
In practice that means three things:
- Many small businesses sit outside the Act because of a turnover test.
- Some are covered whatever their turnover, because of the kind of work they do. Health service providers are the example most owners run into.
- A business can choose to be covered.
Our guide to using ChatGPT safely in a small business sets out the threshold and the main exceptions. Check the Act and the OAIC’s small business pages for your own case before you rely on an exemption.
Even if you are exempt, the principles below are a sensible standard. Larger clients often write privacy obligations into their contracts, and your customers do not know or care what your turnover is when their details leak.
How the Australian Privacy Principles apply to AI tools
The Act does not have a separate set of AI rules. The same principles that cover a filing cabinet or a CRM cover an AI tool. The table below uses the principle headings as they appear in the Act, with the question each one raises when AI is involved.
| Principle | What it covers | The AI question to ask |
|---|---|---|
| APP 1 | Open and transparent management of personal information | Does your privacy policy say you use AI tools, and what for? |
| APP 3 | Collection of solicited personal information | Is a chatbot or call agent collecting more than you need? |
| APP 5 | Notification of the collection of personal information | Do people know an AI tool is taking their details? |
| APP 6 | Use or disclosure of personal information | Is pasting a client file into an AI tool a use for the purpose you collected it for? |
| APP 8 | Cross-border disclosure of personal information | Where does the AI vendor process and store what you send? |
| APP 10 | Quality of personal information | Who checks AI output about a person before you rely on it? |
| APP 11 | Security of personal information | Who can log in to the AI account, and what can they see? |
Two points catch owners out. First, information an AI tool produces about an identifiable person can itself be personal information, so a summary of a phone call or a “lead score” is not outside the rules just because a machine wrote it. Second, the overseas question (APP 8) comes up with most popular AI tools, because the companies behind them are based overseas. Find out where yours processes data before personal information goes in.
The change on 10 December 2026: automated decisions
The Privacy and Other Legislation Amendment Act 2024 received Royal Assent on 10 December 2024. Most of it started the next day. Part 15 of Schedule 1, on automated decisions and privacy policies, was given two years, and the commencement table sets its start date as 10 December 2026.
From then, a covered business’s privacy policy must include extra information if all three of these are true:
- the business has arranged for a computer program to make a decision, or to do something substantially and directly related to making one
- the decision could reasonably be expected to significantly affect an individual’s rights or interests
- personal information about that person is used in the program to make the decision
The policy must then describe the kinds of personal information the programs use, the kinds of decisions made solely by the programs, and the kinds of decisions where a program does something substantially and directly related to the decision.
The Act’s wording is wider than many owners expect. Refusing or failing to make a decision counts as making one. A decision can affect someone’s rights or interests whether the effect is good or bad. The examples the Act gives include a decision that affects a person’s rights under a contract, agreement or arrangement, and a decision that affects their access to a significant service or support.
It also applies to decisions made after 10 December 2026 even if the system was set up, and the information collected, before that date. A tool you installed in 2025 is not grandfathered.
Questions to ask about your own tools
- Does any system approve, decline or rank applications without a person looking first, such as tenancy applications, finance pre-checks or job applicants?
- Does an online quote or booking tool refuse some customers automatically, for example based on location or payment history?
- Does an AI agent decide who gets a callback and who does not?
- Does software change a customer’s price, terms or account status on its own?
A chatbot that answers opening hours is a long way from this. A tool that decides whether someone gets a service is much closer. Where a tool sits in between, get advice on whether the clause applies, and list it anyway so you know what you have.
Other 2024 changes worth knowing
- Security now names “technical and organisational measures”. The amendment added APP 11.3, which says the reasonable steps to protect personal information include technical and organisational measures. For AI tools, that points to things like business accounts instead of personal logins, two-factor sign-in, and clear rules about what staff may paste in.
- A statutory tort for serious invasions of privacy. Schedule 2 gives an individual a cause of action where someone intrudes on their seclusion or misuses information about them, they had a reasonable expectation of privacy, the invasion was intentional or reckless and serious, and the public interest in their privacy outweighs any countervailing public interest. The Act set its start as a day fixed by Proclamation, or six months after Royal Assent if no day was fixed first. There are exemptions, including for journalism.
- Overseas data flows, data breaches and penalties. The same Act also amended the rules on overseas data flows, eligible data breaches and penalties for interference with privacy. Our guide to the Notifiable Data Breaches scheme covers what to do if personal information gets out.
What the OAIC says about AI
The Office of the Australian Information Commissioner, the regulator, publishes guidance on privacy and AI on its website, for businesses using commercial AI products and for those building them. We could not open those pages when writing this article, so we have not summarised them here. Read them directly at oaic.gov.au before you decide how your business will use AI, and check them again from time to time, since regulator guidance is updated more often than the Act.
An AI privacy checklist for Australian small businesses
- List every AI tool in use. Include the free accounts staff signed up for themselves, the AI features switched on inside your CRM or email, and any phone or chat agent.
- For each tool, write down four things: what personal information goes in, where the vendor processes it, whether the vendor’s terms let it use your data to train its models, and who in your business has access.
- Move work onto business accounts so the business, not the staff member, controls the data and can close access when someone leaves.
- Update your privacy policy to say how you use AI. Our guide to a website privacy policy in Australia covers the rest of what belongs in it.
- Tell people when they are dealing with AI, for example at the start of a chatbot conversation or an AI-answered call.
- Map automated decisions before 10 December 2026 using the questions above, and add the required information to your policy if any qualify.
- Keep a person in the loop for anything that affects a customer’s money, access or rights, and check AI output before it reaches a client.
- Write it down for staff. A one-page AI policy stops people guessing.
Firms in professional services and health tend to hold the most sensitive client information, so they have the most to gain from doing this properly.
When CentriWeb sets up AI for a business, such as a phone receptionist or website chat, it answers from the business’s own information and hands over to a person when it should. Working through the questions above is part of that set-up, and so is writing down which decisions the system makes on its own.