Short answer
From 10 December 2026, a business covered by the Privacy Act must say in its privacy policy when it has arranged for a computer program to make a decision about a person, or to do something substantially and directly related to making it, using that person's personal information, where the decision could significantly affect them. The policy has to name the kinds of personal information used and the kinds of decisions involved. It's a rule about what you tell people, not a ban on any tool. For most small practices the job is one sitting with a list of your software and, if anything is in scope, a new section in the privacy policy.
This guide is general information, not legal advice. We're an IT business, not a law firm. Everything below comes from the Act and the regulator's published guidance, which are linked at the bottom and are the documents to rely on.
Is your practice covered at all?
The Office of the Australian Information Commissioner (OAIC) says most small businesses, meaning those with an annual turnover of $3 million or less, are not covered by the Privacy Act. But some are covered whatever they turn over, and the first one on its list is a health service provider. The OAIC's small-business checklist describes that as a business providing services in relation to physical, emotional, psychological and mental health, and it names allied health professionals.
So a physio, OT, psychology or speech practice is covered from its first dollar, and so is the new rule. If you're a small business outside health, run through the OAIC's checklist before you assume you're exempt: trading in personal information, or holding a Commonwealth contract, also brings you in.
What the new rule says
The Privacy and Other Legislation Amendment Act 2024 added three clauses to Australian Privacy Principle 1, numbered APP 1.7 to 1.9. They start on 10 December 2026, two years after the Act became law. Your privacy policy must carry extra information if all three of these are true:
- You have arranged for a computer program to make a decision, or to do a thing that is substantially and directly related to making a decision.
- The decision could reasonably be expected to significantly affect the rights or interests of a person.
- Personal information about that person is used by the program to make the decision or do that thing.
If so, the policy must set out the kinds of personal information the programs use, the kinds of decisions made solely by a program, and the kinds of decisions where a program does something substantially and directly related to making them.
What the words mean, according to the OAIC
The OAIC updated its guidelines on APP 1 on 30 September 2026 and published a fact sheet and a flowchart. The definitions are wider than most people expect.
- Computer program is read broadly. The fact sheet lists pre-programmed rule-based processes, artificial intelligence and machine learning, software, apps and word-processing tools, and generative AI including chatbots. A formula in a spreadsheet counts.
- Arranged for covers buying someone else's software, configuring an off-the-shelf package and relying on a program's advice, as well as anything built in-house. One of the OAIC's own examples is an employer that permits or directs staff to use an AI chat tool to draft performance assessments that determine promotions. The OAIC generally expects the obligation to stay with the business using the personal information to make the decision, not with the software vendor.
- A decision includes refusing or failing to make one, and it counts whether the person ends up better or worse off.
- Substantially and directly related means the program's output is a key factor in the person's decision and has a direct connection with it. The output can be advice; it doesn't have to be the final word.
- Significantly affect means more than trivial, with the potential to considerably influence the person's circumstances. The OAIC's list of interests that are generally in scope includes access to healthcare, access to physical assistance or support, and financial assistance.
A person checking the output doesn't take you out
This is the point most likely to catch a practice. The guidelines say a decision may be in scope even where the program's output doesn't replace the whole decision-making process or is subject to human review. The fact sheet adds that the OAIC considers machine learning or generative AI outputs used to make significant decisions would generally be in scope unless subject to extensive human oversight and control.
What is not captured: a person makes the decision and uses a program for something other than helping to make it. The guidelines' example is using a word processor to document a decision.
Who the decision is about matters
The OAIC says a person's circumstances affect whether a decision is significant. A decision may weigh more heavily on a child, a person with a disability or someone experiencing vulnerability than on other people. A paediatric or NDIS practice should read "significant" with that in mind.
The OAIC's example that looks most like a practice
The first worked example in the fact sheet is a health and aged care provider that keeps a spreadsheet of existing and prospective clients. A formula scores each person's need from details such as date of birth, address and health information, and ranks them. Staff use the ranking to decide who to contact and book in.
The OAIC's conclusion is that this is within the obligation: the formulas determine who is offered health and aged care services, and those services matter enough to affect a person's rights or interests. It adds that using the same ranking to work out who not to contact would be in scope as well.
The fact sheet's list of decisions that would generally be in scope also includes computer programs used to prioritise the provision of health or disability services, recruitment software used to sort candidates and make hiring decisions, and AI-written reports used to rank employee performance or decide promotions and bonuses.
What that looks like in a small practice
This table is our reading of the OAIC's guidance against the tools we see in small practices. It isn't the OAIC's, and it isn't legal advice. Where you're unsure, the OAIC's own advice is to take a cautious approach and include the information.
| What you use | In scope? | Why |
|---|---|---|
| A waitlist spreadsheet, or a feature in your practice software, that scores or ranks who is offered an appointment first | Likely yes | It is the OAIC's own example. Access to healthcare is on its list of significant interests. |
| An online intake or referral form that screens people out by their answers | Likely yes | Refusing is a decision too, and the form uses personal information to do it. |
| A rule in your software that discharges or blocks a client automatically, for example after missed appointments | Possibly | A rule-based process is a computer program. It turns on whether losing the service is significant for that client. |
| An AI tool used to shortlist job applicants or draft staff performance reviews | Likely yes | Both are on the OAIC's lists of examples. |
| An AI scribe that drafts the clinical note for the clinician to review | Probably not, on its own | Writing up a note is not itself a decision about the person. It moves toward scope if what the tool produces becomes a key factor in a decision that matters, such as whether someone keeps receiving a service. Consent and the other privacy rules still apply to it. |
| Appointment reminders, online booking that shows free times, invoice reminders | Probably not | Hard to see a significant effect on anyone's rights or interests. |
| Word, Outlook or your notes system, used by a clinician who makes the decision themselves | No | The guidelines say a program used for something other than facilitating the decision is not captured. |
What the privacy policy has to say
If something is in scope, the policy needs a section that names the kinds of personal information the programs use and the kinds of decisions involved. The OAIC's guidance on how to write it is practical:
- Plain language that a person can understand, without burying them in technical detail.
- You may group similar decisions and kinds of information together, as long as the result still means something to a reasonable person.
- Where more sensitive information is used, and health information is the OAIC's example, say so clearly.
- Set it out so a person can ask for more information, and so the decision can be reviewed or challenged.
- You don't have to include commercial-in-confidence information such as trade secrets.
The fact sheet's two sample policy sections are both headed "Decisions made by computer programs" and are simply a list of the kinds of information followed by a list of the kinds of decisions. For a practice with a ranked waitlist, something along these lines would follow the same pattern. It's an illustration to adapt, not approved wording:
Illustration only
Decisions made by computer programs. We use our practice software to help decide the order in which people on our waiting list are offered appointments. It uses these kinds of personal information: date of birth, referral details, and the health information you or your referrer gave us. A clinician reviews the order before anyone is contacted. To ask how a decision about you was made, or to have it looked at again, contact our practice manager.
A tool you set up years ago still counts
The Act's application clause says the rule applies to decisions made after it starts, whether the arrangement for the program was made before or after that date, and whether the personal information was collected before or after. A waitlist formula written in 2022 is in scope for the decisions it informs from 10 December 2026.
What to do before 10 December
- List the software that touches personal information. Practice management, booking, intake forms, spreadsheets with formulas, recruitment tools, and any AI tool staff use, including ones they signed up for themselves.
- Put each one through the OAIC's flowchart. Does it use personal information? Have you arranged for it to make or inform a decision? Is it substantially and directly related to that decision? Could the decision significantly affect the person?
- Ask your vendors. The OAIC says providers of third-party software should give clear, high-level information about how their software can be used to make decisions, so that you can describe it in your policy. Ask in writing.
- Write the section and publish it. APP 1 already requires the policy to be up to date and available free of charge, usually on your website.
- Tell staff, and keep the list. Go back to it whenever a tool is added or switched on. A new AI feature in software you already own can change the answer.
If nothing you use is in scope, this rule doesn't require a new section. We'd still keep a dated note of what you checked, because that's the first thing you'll be asked for if the question ever comes up.
If you're an NDIS provider
The NDIS Quality and Safeguards Commission published a position statement in February 2026 on using AI to develop behaviour support plans. It says the Commission does not endorse or approve AI tools for developing and reviewing those plans, and that this doesn't prevent providers using AI so long as the use complies with their legal obligations. If a provider does use AI, the Commission expects all information to be appropriately de-identified and no personal information of participants to be disclosed to AI systems.
Among the risks it lists is automated decision-making without appropriate human oversight or clinical judgement. It also says that disclosing sensitive personal information to a third-party AI platform without appropriate safeguards may breach the NDIS Code of Conduct's requirement to respect the privacy of people with disability. The statement is about behaviour support plans specifically, but it tells you how the Commission is thinking. Our page on IT for NDIS providers covers the systems side.
The AI rules that already apply
None of this replaces what was already expected. The OAIC's guidance on commercially available AI products says privacy obligations apply to any personal information put into an AI system and to what it produces, that businesses should update their privacy policies and notices with clear information about their use of AI, and that as a matter of best practice you should not enter personal information, and particularly sensitive information, into publicly available generative AI tools. Our guide to which AI tools are safe to use at work turns that into accounts, a never-paste list and a one-page policy.
What may come next
The 10 December rule is law. The next round is not, yet. The Attorney-General's Department consulted on an exposure draft of the Privacy Amendment (Personal Data Protection) Bill 2026 from 31 August to 18 September 2026, and says the Bill remains subject to further consideration by government. The draft's commencement dates are blank. As drafted, it would:
- require every collection, use or disclosure of personal information to be fair and reasonable in the circumstances, as well as lawful;
- say that consent must be voluntary, informed, current, specific and unambiguous;
- require a statement to the Commissioner within 72 hours once a business has reasonable grounds to believe there has been an eligible data breach;
- add a right to have personal information destroyed on request, which the draft applies to large digital platforms, not to a practice.
None of that needs action today, and it may change before it reaches Parliament. The one thing worth doing now costs nothing: decide who in the practice would write a breach statement, and where the facts they'd need are kept.
Where we fit
For the health and NDIS practices we look after, the software list in step one is something we already hold, because we manage the systems. We go through it against the OAIC's flowchart with you, ask the vendors the questions, and word the policy section in plain English as part of our security and compliance work. Where something is borderline we say so and tell you it needs a lawyer's eye. More on how we work with practices is on our IT for allied health page.
Sources
- OAIC: New resources on transparency for use of AI and automated decision-making (30 September 2026), APP Guidelines, Chapter 1: APP 1 (version 2.0), APP 1.7-1.9 transparency obligation fact sheet (PDF) and flowchart (PDF)
- OAIC: Small business (who the Privacy Act covers, and the checklist), Guidance on privacy and the use of commercially available AI products
- Federal Register of Legislation: Privacy and Other Legislation Amendment Act 2024 (Schedule 1, Part 15: the new clauses, the start date and the application clause)
- Attorney-General's Department: Privacy Reform: consultation on exposure draft legislation and the exposure draft Bill (PDF)
- NDIS Quality and Safeguards Commission: Position statement: use of artificial intelligence in the development of behaviour support plans (PDF, February 2026)
Checked 15 November 2026. This is general information, not legal advice; the Act and the OAIC's pages are the ones to rely on, and a lawyer is the person to ask about your own practice. If something here looks out of date, tell us.