Skip to content
Guides

Automated decisions and your privacy policy: what changes on 10 December 2026

Vetrox · Published

The short answer

From 10 December 2026, if the Privacy Act covers your business and a computer program, AI included, uses someone's personal information to make a decision that could significantly affect them, or to do something substantially and directly related to making it, your privacy policy has to say so. It must set out the kinds of personal information those programs use, the kinds of decisions they make on their own and the kinds they help a person make. Most of the work is finding those programs: the systems, spreadsheets and AI tools that score, rank, approve or turn people down.

What changes on 10 December 2026

The Privacy and Other Legislation Amendment Act 2024 (opens in a new tab) adds a new requirement to Australian Privacy Principle (APP) 1, the principle that requires a privacy policy. It takes effect on 10 December 2026, two years after the Act received Royal Assent. From then, APP 1.7 says a privacy policy must include more information when 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 it
  • the decision could reasonably be expected to significantly affect the rights or interests of an individual
  • personal information about that individual is used when the program runs.

The rule applies to decisions made after it takes effect, even when the program was set up, or the personal information collected, before then. A system that has scored applications for years is covered from the first day.

On 30 September 2026 the Office of the Australian Information Commissioner (OAIC) published its guidance (opens in a new tab): an updated chapter of its APP guidelines, a fact sheet (with a second one for government agencies) and a flowchart. This guide follows them. Under the same Act, a privacy policy that leaves the new information out can lead to a compliance notice or an infringement notice, as one that leaves out the other required information already can.

Does it apply to your business?

The Australian Privacy Principles only bind businesses the Privacy Act covers, so start there. Most small businesses are not covered, but some are (opens in a new tab): health service providers, businesses that trade in personal information and contractors providing services under a Commonwealth contract, for example, are covered whatever their turnover. The OAIC's page has a checklist for working it out.

If the Act covers you, the OAIC's flowchart (opens in a new tab) works through the rest in order:

Does the new rule apply? Four questions, in order

  1. 1

    Does the Privacy Act cover your business?

    Most small businesses are not covered, but some are

  2. 2

    Does a computer program make or inform a decision about a person, using their personal information?

    Including software you bought or set up, spreadsheets and AI tools

  3. 3

    Is the program a key factor, directly connected to the decision?

    Even when a person makes the final call

  4. 4

    Could the decision significantly affect the person?

    More than trivial, such as a job, a loan, insurance or care

Based on the OAIC's flowchart. A clear no at any step puts the decision outside the rule; yes to all four means your privacy policy needs the new information.

When you are not sure, the OAIC's fact sheet (opens in a new tab) says to take a cautious approach and include the information in your privacy policy. The next two sections explain the questions in the middle.

What counts as a computer program, and as a decision

Both terms are read broadly. The OAIC's APP 1 guidelines (opens in a new tab) say a computer program includes pre-programmed rules, artificial intelligence and machine learning, as well as everyday software, apps and word-processing tools, and that generative AI tools, chatbots included, all count.

It does not have to be AI. In one of the fact sheet's examples, a health and aged care provider uses a formula in a spreadsheet to score clients and rank them by need, and staff choose whom to contact from the ranking. That spreadsheet is in scope, and it would still be if staff used the ranking to decide whom not to contact.

A decision includes refusing or failing to decide or to act, and it counts whether the effect on the person is good or bad. A program that only recommends can still be in scope: the OAIC calls it substantially related when it is a key factor in a person's decision, and directly related when it has a direct connection with making it. A decision can be in scope even when the program does not replace the whole process, or a person reviews its output.

In another example, managers use a generative AI tool that suggests each salesperson's bonus and pay rise. Every final decision needs a director's written approval, but the managers use the suggestions as a key factor, so the tool's output counts as substantially and directly related to the decision. The OAIC adds that questioning the output, reviewing the performance measures behind it, using other evidence and recording why a decision departs from the suggestion could indicate the output was not substantially related.

For machine learning and generative AI, the OAIC's view is that outputs used in decisions that significantly affect people will generally be in scope unless there is extensive human oversight and control. A word processor used only to write up a decision a person has already made is not in scope.

Which decisions significantly affect someone

The effect has to be more than trivial and able to considerably influence the person's circumstances or outcomes. The OAIC's fact sheet lists decisions it would generally treat as in scope, including:

  • recruitment software that sorts candidates and makes hiring decisions
  • AI reports used to rank staff performance or decide promotions, bonuses or pay
  • approving or rejecting a loan or credit application
  • assessing whether someone is eligible for insurance
  • prioritising health or disability services
  • online retailers setting different prices for different customers on significant goods
  • facial recognition in a shop or stadium that matches people against a watchlist.

The person's situation matters too. The guidelines' example is a rideshare service whose software can suspend accounts: a suspension may matter little to most customers, but could significantly affect a customer experiencing vulnerability who relies on the service to get to medical appointments or keep their job. The guidelines also list areas generally treated as significant, among them employment, housing, healthcare and aged care, education, banking and credit, telecommunications and essential utilities.

Software you buy counts too

The rule falls on the business that has arranged for a program to make or inform the decision, whoever wrote the program. The fact sheet's examples of arranging for a program include building one in-house, buying one from another company, configuring or customising off-the-shelf software, authorising a program to make a decision and relying on a program's advice.

The guidelines add everyday cases: an employer that lets staff use, or tells them to use, an AI chat tool to draft performance assessments that decide promotions; a business that has a software company approve or decline refunds automatically on its behalf; and a case management system that escalates certain complaints on its own.

The OAIC expects the obligation to stay with the business that uses the personal information to make the decision, even when another company provides or runs the program, and a contract with that company may not change it. It asks businesses to understand how the products they buy make or assist decisions, both when buying and afterwards, and to make clear in the contract which side keeps the final decision. Vendors, in turn, should give their customers clear, high-level information about how their software makes decisions, for the customers' privacy policies.

What the privacy policy must say

When the rule applies, APP 1.8 requires your privacy policy to set out:

  • the kinds of personal information those programs use
  • the kinds of decisions the programs make on their own
  • the kinds of decisions where a program does something substantially and directly related to making them, such as recommending an outcome to a person.

This sits alongside what APP 1.4 already requires in every privacy policy, such as the kinds of personal information you collect and hold, what you use it for and how people can access it or complain.

You can group decisions and kinds of information, but the OAIC says the detail must be meaningful to a reasonable person: plain language, specific to your business, and framed so a decision can be challenged. Sensitive information, such as health information or biometric templates, should be named clearly. Commercially sensitive information and trade secrets, such as how a program weighs each piece of information, can stay out, but the OAIC reads that exclusion narrowly: using a program to make decisions about people is not, in itself, commercial-in-confidence.

The fact sheet includes sample wording: a lender adds a section headed “Decisions made by computer programs” that lists the kinds of personal information its fraud checks use, such as identity information, location data, biometric templates and credit information, and the decisions they lead to, such as granting or denying a transaction or a loan application.

A privacy policy is a legal document, so have your lawyer or privacy adviser write or check the new section. What they need from you is the list in the next section.

What to do before 10 December

Most of the work is finding the programs and writing down what they do. In order:

  1. Check whether the Privacy Act covers your business, with the OAIC's checklist or your adviser.
  2. List every program that makes or informs a decision about a person: your own systems, software you subscribe to, automations, spreadsheets with scoring formulas and the AI tools staff use, for screening applicants or assessing performance, for example.
  3. For each one, write down the decision, the personal information it uses, whether it decides on its own or a person decides with its output, and how much that person relies on it.
  4. Mark the decisions that could significantly affect someone, such as hiring, pay and promotion, credit, insurance, housing or care.
  5. Ask each vendor for clear, high-level information about how their product makes or assists decisions and which personal information it uses.
  6. Have your lawyer or privacy adviser write or check the new section, and publish the updated privacy policy before 10 December 2026.
  7. Keep the list up to date as you add systems and AI tools, and review the policy regularly. The OAIC suggests doing it at least as part of your annual planning, and showing on the policy when it was last updated.

While you are listing them, look at how each decision is made, not only at how to describe it. A decision is easier to explain and review when your systems record what the program used and what it returned, and a person can question the output.

Where Vetrox fits

Vetrox is an Australian IT company that builds AI and automation into the systems businesses across Australia already use. In what we build, a log keeps what the AI read and what it returned, and an item waits for a person to approve it when the AI is unsure or a wrong answer would cost you. For this change, we can go through your systems, automations and AI tools with you and write down which decisions each one makes or informs and which personal information it uses, so whoever writes your privacy policy works from facts. The wording of the policy itself is for your lawyer or privacy adviser.

Sources

  1. Privacy and Other Legislation Amendment Act 2024 (Cth), as made: Schedule 1, Part 15, Automated decisions and privacy policies (opens in a new tab)
  2. OAIC: New resources on transparency for use of AI and automated decision-making (published 30 September 2026) (opens in a new tab)
  3. OAIC: Small business (which businesses the Privacy Act covers) (opens in a new tab)
  4. OAIC: APP 1.7-1.9 Transparency Obligation Flowchart (opens in a new tab)
  5. OAIC: APP 1.7-1.9 Transparency Obligation Fact Sheet (September 2026) (opens in a new tab)
  6. OAIC: Australian Privacy Principles guidelines, Chapter 1: APP 1, Open and transparent management of personal information (updated 30 September 2026) (opens in a new tab)

Show us how your business runs today.

A 30-minute call to discuss the software you use, the work it handles and what needs to change.