·10 min read

Privacy Act Automated Decision-Making: What Your Business Must Disclose From 10 December 2026

From 10 December 2026, APP entities must describe in their privacy policy any computer program that makes or substantially informs significant decisions about people. Who is covered, what counts and what to write.

From 10 December 2026, if your business is covered by the Privacy Act and it has arranged for a computer program to use personal information to make, or do something substantially and directly related to making, a decision that could reasonably be expected to significantly affect a person's rights or interests, your privacy policy must say so. It must describe the kinds of personal information the program uses and the kinds of decisions involved. The rule is in new Australian Privacy Principles 1.7 to 1.9, inserted by the Privacy and Other Legislation Amendment Act 2024, and the Office of the Australian Information Commissioner (OAIC) published its final guidance on 30 September 2026. It is a transparency obligation about what your privacy policy says, not a ban on automated decisions.

This guide is general information, not legal advice. It walks through who the rule applies to, how the OAIC reads each part of the test, how common business automations measure up against it, and what a compliant privacy policy section can look like. For anything borderline, read the OAIC material linked below and get advice from a privacy lawyer.

Does this apply to your business?

The obligation applies to APP entities: the organisations and agencies the Privacy Act covers. The OAIC's small business guidance states that most small businesses are not covered, and that a small business for this purpose is one with an annual turnover of $3 million or less, counting income from all sources.

The same OAIC page lists businesses that are covered regardless of turnover. They include:

  • health service providers, which the OAIC says includes medical practitioners, pharmacists, allied health professionals, complementary therapists, child care centres and private schools
  • businesses that trade in personal information
  • contractors providing services under a Commonwealth contract
  • credit reporting bodies and reporting entities under the Anti-Money Laundering and Counter-Terrorism Financing Act 2006
  • businesses related to a business the Privacy Act covers
  • businesses that have opted in to the Privacy Act.

If none of those apply and your turnover has stayed at or under $3 million, the APPs, including this new disclosure, do not currently apply to you. If you are an APP entity, read on.

The three-part test, in plain English

The OAIC's APP Guidelines, Chapter 1 (paragraph 1.35) set out when APP 1.7 is triggered. All three parts must be met.

1. You have arranged for a computer program to make the decision, or to do something substantially and directly related to making it

"Computer program" is read broadly. The OAIC's APP 1.7–1.9 fact sheet says it includes pre-programmed rule-based processes, artificial intelligence and machine learning, ordinary software, apps and spreadsheets, and generative AI including chatbots. One of the OAIC's own examples is a spreadsheet formula that ranks clients for follow-up.

"Arranged for" covers software you built, software you bought, off-the-shelf software you configured, and advisory outputs your staff rely on. The OAIC lists as examples an employer permitting staff to use an AI chat tool to draft performance assessments that determine promotions, and a business contracting a software company to approve or decline refunds automatically (APP Guidelines, paragraphs 1.42 to 1.45).

"Substantially and directly related" catches programs that recommend or guide rather than decide. The OAIC defines substantially as "a key factor in facilitating the human's decision making" and directly as having "a direct connection with making the decision" (paragraph 1.50). The fact sheet adds that machine learning or generative AI outputs used to make significant decisions would generally be in scope "unless subject to extensive human oversight and control". Using a word processor to write up a decision a person has already made is not captured (paragraph 1.40).

2. The decision could reasonably be expected to significantly affect someone's rights or interests

The OAIC says significantly means the effect must be "more than trivial" and must have "the potential to considerably influence the circumstances or outcomes for the individual concerned" (paragraph 1.57). A decision counts whether its effect is good or bad for the person (APP 1.9), and refusing or failing to decide counts as a decision.

The OAIC's examples of interests include employment, financial, health, housing and access to significant goods and services (paragraphs 1.61 to 1.66). It also says vulnerability matters: a decision with a minor effect on most people can be significant for a child or a person experiencing vulnerability.

3. Personal information about the person is used in the program's operation

If the program does not use personal information about the individual the decision is about, APP 1.7 is not triggered. The OAIC's flowchart notes that computer programs are only in scope "to the extent that an APP entity uses personal information to make or inform a decision".

Common business automations against the test

The table below applies the OAIC's test and examples to automations that many businesses run. It is a starting point for your own review, not a ruling. Where the OAIC has given a close example, it is noted. Where you are unsure, the OAIC fact sheet says to "take a cautious approach and include information" in your privacy policy.

AutomationLikely positionWhy
Software or an AI tool that screens, scores or ranks job applicantsLikely in scopeThe OAIC lists recruitment software that sorts candidates and makes hiring decisions as generally in scope (fact sheet, Note 1).
An AI tool that drafts performance reviews, bonuses or pay recommendations that managers rely onLikely in scopeOAIC fact sheet Example 3 and APP Guidelines paragraph 1.45. Human approval does not take it out of scope if the output is a key factor.
A rule that automatically puts a customer's account on credit hold or declines creditLikely in scopeApproving or rejecting credit is on the OAIC's list of in-scope decisions, and decisions affecting rights under a contract are given as examples (paragraph 1.63).
A chatbot or workflow that approves or declines refundsLikely in scope where the refund matters to the customerContracting software to approve or decline refunds is an OAIC "arranged for" example. Whether the effect is significant depends on the amount and the customer.
Pricing that changes by customer, postcode or behaviourLikely in scope for significant goods or servicesOAIC fact sheet Example 4 treats differential pricing on essential goods as significant even when individual differences are small.
A spreadsheet or system that triages or prioritises clients for health, care or support servicesLikely in scopeOAIC fact sheet Example 1 is a spreadsheet formula used this way.
Rostering software that allocates shifts, and so hours and payPossibly in scopeEmployment and financial interests are listed by the OAIC (paragraph 1.61). Assess how much the output decides and how much a person reviews it.
Automatic lead scoring that decides which enquiries get a call backPossibly in scopeDepends on whether missing out could considerably affect the person. A sales follow-up order rarely will; access to an essential service could.
An AI tool that drafts an email reply a staff member rewrites before sendingLikely out of scopeNo decision about the person is being made or substantially informed, provided the tool is not deciding the outcome of the request.
Scheduled newsletters and appointment remindersLikely out of scopeThese do not decide anything about the person's rights or interests. Targeted advertising can be in scope where it leads to personalised pricing for significant goods or limits access to employment (paragraph 1.63).

For more on where automation is commonly used in Australian businesses, see our guide to AI automation examples for Australian small businesses. Several of those examples involve a decision about a person and would need this review.

What to write in your privacy policy

APP 1.8, as summarised in the OAIC's APP Guidelines (paragraph 1.36), requires three pieces of information:

  1. the kinds of personal information used in the operation of the computer programs
  2. the kinds of decisions made solely by the operation of the computer programs
  3. the kinds of decisions for which the computer programs do something substantially and directly related to making the decision.

The OAIC fact sheet says you may group decisions and information into categories, but the detail "must be meaningful to a reasonable person", and the use of more sensitive information, such as health information or biometric templates, should be clear. It also says to avoid overwhelming readers with technical detail and to frame the information so people can request more and the decision can be reviewed.

Commercially sensitive information and trade secrets about your computer programs are excluded, so you do not need to publish how a model weights its inputs. The OAIC is clear that the exclusion is narrow: the fact that you use personal information to make a kind of decision is not commercial-in-confidence, and embarrassment or likely criticism is not a reason to leave it out.

A worked example section

The OAIC's fact sheet includes two sample privacy policy sections (Examples 6 and 7). The structure below follows them. The business, programs and decisions are illustrative. Replace them with what your inventory finds.

Decisions made with the help of computer programs

We use computer programs to help us assess job applications. These programs use the following kinds of personal information: your name and contact details, the information in your resume and application, and your answers to screening questions. They are used to rank applications and recommend which applicants to invite to interview. A member of our team reviews the recommendations and makes the final decision about who is interviewed.

We use computer programs to manage customer accounts. These programs use the following kinds of personal information: your name, contact details, account and payment history. They are used to make the following kinds of decisions: placing an account on hold when payments are overdue, and approving or declining refund requests under a set amount.

If you would like more information about how these decisions are made, contact our privacy officer at [contact details].

Note the difference between the two paragraphs. The first describes decisions a program substantially informs. The second describes decisions a program makes on its own. APP 1.8 asks for both kinds, described separately.

If you use third-party software

The OAIC generally expects the disclosure obligation to remain with the business that uses personal information to make the decision, even when the program belongs to a vendor (fact sheet; APP Guidelines paragraph 1.44). It says businesses should "actively identify, assess and keep oversight over the third-party computer programs that they use to make a decision, both during and after procurement" (paragraph 1.45), and that vendors should provide clear, high-level information about how their software can be used to make decisions.

In practice that means asking each vendor whose product scores, ranks, approves, declines or prioritises people:

  • which personal information the product uses for each of those functions
  • whether the product decides or recommends, and which settings change that
  • which of those functions you have switched on.

Keep the answers with your inventory. The OAIC's APP Guidelines also say contracts should be clear about which party keeps ultimate decision-making capability (paragraph 1.45).

Automations built before the deadline still count

Item 89 of Schedule 1 to the Privacy and Other Legislation Amendment Act 2024 applies the new principles to decisions made after commencement, whether the arrangement for the computer program was made before or after commencement, and whether the personal information was acquired or created before or after it. The OAIC's APP 1 Guidelines say the same: the amendments apply to decisions made from 10 December 2026 "regardless of whether the arrangement for a computer program to make the decision was made before or after this date". An approval flow set up three years ago that is still running needs to be in the inventory.

Why it is worth getting right

Item 87 of the same Part adds APP 1.7 to section 13K of the Privacy Act, the list of civil penalty provisions for which infringement notices or compliance notices can be issued. A privacy policy that leaves out an in-scope automated decision is therefore not only a transparency gap: it is a contravention for which an infringement notice or compliance notice can be issued.

A checklist to finish before 10 December 2026

  1. Confirm whether you are an APP entity. Check your turnover and the OAIC's list of businesses covered regardless of turnover.
  2. List every program that touches a decision about a person. Include spreadsheets with formulas, rules in your accounting or booking software, automation workflows, chatbots, and AI tools staff use to draft assessments.
  3. For each one, apply the three parts of the test. Does it make or substantially and directly inform a decision? Could the decision significantly affect someone? Does it use their personal information?
  4. Ask your vendors. Get the kinds of personal information used and whether each function decides or recommends.
  5. Write the privacy policy section. Separate decisions made solely by a program from decisions a program substantially informs. Make any use of sensitive information clear.
  6. Decide how people can ask for more information. The OAIC suggests structuring the section so people can request further detail.
  7. Get it reviewed. Have a privacy lawyer, or at least someone who did not write it, check the section against the OAIC fact sheet.
  8. Keep the inventory current. Add a step to your change process so a new automation, or a new AI tool used for assessments, triggers a privacy policy check.

The inventory is also the foundation of a wider AI governance approach. Our AI adoption and governance framework covers policy, approved tools and review in more detail.

Where Nublock fits

Nublock Digital, the trading name of Nublock Group Pty Ltd, is a Melbourne-based business that provides AI consulting and automation alongside managed IT. For AI builds, the client owns the code, prompts and accounts, so the automations themselves stay with the business that has to disclose them. If you would like help building your inventory of automations, get in touch. For the legal wording of your privacy policy, use a privacy lawyer.

Sources

Does the automated decision-making rule apply to small businesses?

Only if the business is covered by the Privacy Act. The OAIC explains that most businesses with an annual turnover of $3 million or less are not covered, but some are regardless of turnover, including health service providers, businesses that trade in personal information, Commonwealth contracted service providers, businesses related to a covered business, and businesses that have opted in. If your business is an APP entity, the new APP 1.7 to 1.9 disclosure applies to you from 10 December 2026.

Does a tool count if a person makes the final decision?

It can. The OAIC says a decision may be in scope even where the computer program does not replace the whole decision-making process or is subject to human review. What matters is whether the program does something substantially and directly related to the decision, meaning it is a key factor in the human decision and has a direct connection to it. The OAIC also says machine learning or generative AI outputs used to make significant decisions would generally be in scope unless they are subject to extensive human oversight and control.

What exactly has to go in the privacy policy?

APP 1.8 requires three things: the kinds of personal information used in the operation of the computer programs, the kinds of decisions made solely by the operation of those programs, and the kinds of decisions for which the programs do something substantially and directly related to making the decision. The OAIC allows these to be grouped into categories, as long as the detail is meaningful to a reasonable person and any use of more sensitive information, such as health information or biometric templates, is made clear.

Do we have to explain how our software or our vendor's algorithm works?

No. The OAIC says commercially sensitive information and trade secrets about computer programs are excluded, so you do not have to publish how a model weights its inputs. The exclusion is narrow: the fact that you use personal information to make a particular kind of decision is not commercial-in-confidence. If you use a third-party product, the OAIC generally expects the disclosure to remain your responsibility, using information your vendor provides.

Does it apply to automations we set up before 10 December 2026?

Yes. According to the OAIC, the obligation applies to decisions made from 10 December 2026, regardless of whether the arrangement for the computer program was made before or after that date, and regardless of when the personal information was collected or created. An automation built years ago that is still making or informing significant decisions after that date needs to be reflected in your privacy policy.