Does AI Compliance Really Matter in 2026? What the 10 December Privacy Act Deadline Means for Your Business
Back to Latest News
AI & PrivacyPrivacy ActAPP 1.7–1.9OAICAI ComplianceAutomated Decision-MakingAustralian SMEs

Does AI Compliance Really Matter in 2026? What the 10 December Privacy Act Deadline Means for Your Business

On 10 December 2026, new privacy policy transparency obligations in APPs 1.7–1.9 commence. The OAIC’s final guidance answers some important questions that can help Australian SMEs cut through the noise.

Evolve with AI15 November 2026
Does AI Compliance Really Matter in 2026? What the 10 December Privacy Act Deadline Means for Your Business

Important: This article provides general information only and is not legal or privacy advice. It does not take your circumstances into account, and reading it does not create an advisory relationship with Evolve With AI. Privacy obligations depend on your circumstances, so speak with a qualified privacy professional about your business.

If you’re an Australian SME owner, you may already feel buried in software, policies and AI headlines.

Now there’s a date to add to the list.

On 10 December 2026, new privacy policy transparency obligations in Australian Privacy Principles (APPs) 1.7–1.9 commence. As at 16 November, that’s just 24 days away.

The OAIC published its final guidance on 30 September 2026. It includes an updated Chapter 1 of the APP Guidelines, a fact sheet and a flowchart. If you haven’t read them, you’re not alone.

But the guidance answers some important questions that can help you cut through the noise.

Guidance of this kind is not legislation, but it shows how the regulator is likely to interpret the obligation. That is why it is worth reading before 10 December.

First, what is changing?

This is not a new “Privacy Act 2026”. The law remains the Privacy Act 1988, as amended by the Privacy and Other Legislation Amendment Act 2024.

From 10 December, an APP entity’s privacy policy must include information about certain decisions made, or assisted, by computer programs. The obligation applies when all three of these conditions are met:

  1. You have arranged for a computer program to make a decision, or do something substantially and directly related to making it.
  2. The decision could reasonably be expected to significantly affect an individual’s rights or interests.
  3. The person’s personal information is used in the program’s operation.

The decision can affect someone positively or negatively. And “making a decision” includes refusing or failing to make one.

The focus is transparency, not a ban on AI.

What the OAIC guidance clarifies

The final guidance matters because it makes clear that this obligation is not limited to businesses building their own AI systems.

Illustration of a decision flow branching through a person to a shield, representing automated decisions that use personal information

1. “Computer program” means more than AI

The OAIC interprets “computer program” broadly. It can include pre-programmed rules, machine learning, ordinary software, apps, word-processing tools and generative AI, including chatbots.

A practical example: a spreadsheet formula that scores and ranks aged care clients can be in scope if the other parts of the test are met.

So don’t just search your business for the letters “AI”.

2. A human in the loop is not an automatic exit

A decision can still be in scope when a program’s output informs only part of the decision, or when someone reviews that output.

The guidance says machine learning or generative AI outputs used for significant decisions would generally be in scope unless there is extensive human oversight and control.

A quick approval click after seeing a score may not amount to meaningful review.

3. Buying software can count as arranging it

You don’t need to build the program yourself. Procuring a third-party tool, configuring off-the-shelf software or relying on its advisory output can count as arranging for a program to operate.

The obligation typically remains with the business using the program, rather than the vendor. Ask suppliers for enough information to understand what their tool does with personal information and decisions.

4. Small differences can still add up

In the OAIC’s own worked example, postcode-based pricing that added around $132 a year to the cost of essential goods, a matter of a few dollars per item, was significant, because the impact compounds across repeated transactions.

A modest change in one transaction may not look significant in isolation. A repeated or cumulative effect may tell a different story.

5. You can protect sensitive model logic, but disclose the use

APP 1.9 allows certain commercially sensitive information and trade secrets to stay out of the privacy policy. That does not mean you can hide the fact that personal information is used in relevant computer-program decision-making.

You generally don’t need to publish a model’s secret formula. You do need to describe the relevant kinds of information and decisions in your policy.

The exclusion is also narrow in a second way. You can’t withhold information simply because publishing it might be embarrassing or attract criticism. The test is commercial sensitivity, not reputational comfort.

When in doubt, disclose at a meaningful level and get advice.

Where it hides in ordinary SME tools

Start with decisions, not with a hunt for fancy AI. Look at systems that handle:

  • Pricing rules, including location-based pricing
  • Refund, complaint and insurance claim flows
  • Eligibility checks for services, discounts or support
  • Recruitment software that sorts or ranks candidates
  • Reports that prioritise customers, applications or cases
  • Support chatbots that can approve requests
  • Advertising tools that target or exclude people

Not every automated task is covered. The decision must meet the three-part test. But routine software deserves a closer look when it uses personal information and can affect someone’s rights or interests.

What meaningful human review looks like

If your team reviews an automated recommendation, make that review real and recordable.

  • Show the reviewer the inputs and reasons, not just a score.
  • Ask for a short reason when the reviewer departs from the suggestion.
  • Store the program’s output and the final decision side by side.
  • Monitor the override rate. If reviewers almost never disagree, the program may effectively be making the decision.

A human name on the final approval does not tell the whole story.

Your countdown plan to 10 December

Illustration of connected checklist cards with ticks, representing a step-by-step compliance countdown plan

Phase 1: Find the decision points, now

List systems that recommend, rank, approve, refuse, price, prioritise or route decisions about individuals. Include spreadsheets, vendor platforms and ordinary business software.

Phase 2: Apply the three-part test, next

For each decision, ask:

  1. Have we arranged for a computer program to make or substantially and directly assist with this decision?
  2. Could the decision reasonably be expected to significantly affect a person’s rights or interests?
  3. Is personal information about that person used by the program?

Record your reasoning. If you’re unsure, mark it for professional advice rather than assuming it’s out of scope.

Phase 3: Ask vendors in writing, before the end of November

Ask what personal information the tool uses, what kinds of decisions it makes or assists with, and whether human review is built into the process. Keep the answers with your decision records.

Phase 4: Draft, review and publish, by 10 December

Prepare a plain-language privacy policy section, check it against your actual systems, get appropriate review, then publish it before the obligations commence.

What should the policy section say?

Use a clear heading, such as “Decisions made by computer programs”. Explain, in sensible groups:

  • The kinds of personal information used
  • The kinds of decisions made solely by programs
  • The kinds of decisions programs substantially and directly assist with
  • Whether sensitive information is involved
  • How people can contact you to ask for more detail

Avoid vague lines like “we may use AI”. That tells people very little.

A clearer approach might describe, for example, that customer contact and transaction information may be used in automated eligibility checks, and explain the kinds of services those checks relate to. Tailor the wording to your actual practices and seek advice on what is appropriate for your circumstances.

Check whether the Privacy Act applies to you

Not every SME is automatically covered. A business with annual turnover of $3 million or less may be exempt from the Privacy Act’s APP requirements.

But the small business exemption has exceptions. These can include health service providers, businesses trading in personal information, certain Commonwealth contractors and credit reporting bodies, among others. The exemption’s application is fact-specific, so don’t assume your turnover alone settles the question.

If you’re unsure whether you’re an APP entity, check the OAIC’s small business guidance and speak with a qualified privacy professional.

A quick self-check

  • ⚠️ Does any tool rank, approve, refuse or price something for an individual?
  • ⚠️ Does it use personal information to do so?
  • ⚠️ Could the outcome meaningfully affect someone’s rights or interests?
  • ⚠️ Are you relying on a vendor’s assurance without understanding the process?
  • ⚠️ Can your reviewers explain the reasons behind their final decisions?

A “yes” or “not sure” is a useful prompt to investigate, not a conclusion that you’ve breached the law.

The work doesn’t end on 10 December

Every new pricing rule, model, vendor or automation can change what your privacy policy needs to say.

Keep a decision register with the tool, purpose, personal information used, decision affected, human review arrangements and the date you last checked the policy. Update it when the system or its use changes.

That’s more reliable than trying to reconstruct your setup once a year.

If you’re sorting through AI or automation decisions and want a clearer starting point, explore our AI readiness assessment or AI strategy roadmap. If you’d like to talk, book a session: your first Discovery Audit session and your first 15-minute discovery call are each free, with no obligation, no minimum spend and no credit card required.

Sources


Disclaimer: This article provides general information only and does not constitute legal, privacy, financial or other professional advice. It does not take your objectives, circumstances, systems or data into account. Reading it does not create an advisory relationship with Evolve With AI. While care has been taken in preparing it, no warranty is given as to its accuracy, completeness or suitability for your circumstances, and we make no representation that any particular approach will make your business compliant. Complying with the APP 1.7–1.9 transparency obligation does not by itself make an organisation compliant with the Privacy Act 1988 (Cth) or any other law. Information is current as at 14 November 2026 and may change; the OAIC’s guidance may also be updated. AI-assisted drafting was used. Verify the information and seek qualified advice before acting on it.

Ready to Start Your AI Transformation?

Book a free discovery call and let's explore how AI can transform your business.